Author SHA1 Message Date
Alexey Bodanyuk b6847e2e68 docs: адрес форка в REPORT 2026-10-20 03:22:03 +03:00
Alexey Bodanyuk 49482529ec docs: https://gitea.privcloud.tech/wfa/mtusi2026lab3 2026-10-20 03:06:32 +03:00
Alexey BodanyukandClaude Opus 5.5 d0dedee773 docs: числа отказов по доказательному прогону (pause: 4 попытки, <=3,2 с)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 18:29:03 +03:00
Alexey BodanyukandClaude Opus 5.5 c758e49cbb docs: REPORT lab3 — доказательства со стенда из чистого clone, уровень 4, трек 1
Второй прогон scripts/collect_evidence.py на новом clone (коммит 2708732):
make check/validate/test, test-persistence (новые PID у lb, api x2,
postgres, valkey), test-failover для обеих реплик по очереди (0 неудачных
попыток), 201 свой тест, серия отказов stop/kill/pause x 2 реплики,
rolling update под нагрузкой (0 ошибок), tests/load.py. Таблица
критерий -> команда -> результат -> файл, раздел уровня 4 с периодами
проверок и временем обнаружения.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 18:28:25 +03:00
Alexey BodanyukandClaude Opus 5.5 27087326d0 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
Найдено первым прогоном из чистого clone: make test-failover (BASE_URL
по умолчанию http://localhost:8080) успел сделать 8 записей за 15 с, и
первая успешная запись после хука пришла на 8,4 с. Windows сначала
стучится на ::1, где порт не был опубликован, и каждое новое соединение
ждёт ~2 с. LB теперь опубликован на 127.0.0.1 и [::1] (оба loopback):
0,013 с на запрос. Тест политики портов допускает только loopback.

collect_evidence: публичный тест всегда выбирает реплику, обслужившую
его первый POST, а последовательность запросов фиксирована — перед
каждым следующим прогоном rotation сдвигается лишними GET, чтобы
по очереди останавливались обе реплики.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 18:09:12 +03:00
Alexey BodanyukandClaude Opus 5.5 2846cae118 fix: collect_evidence передаёт make интерпретатор без пробелов в пути
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 17:34:43 +03:00
Alexey BodanyukandClaude Opus 5.5 b9f1ad58f9 bench: трек 1 — профили py-spy до/после, серия E (потолок нового пути), графики
Профиль при 240 req/s: доля psycopg 52.7 % -> 19.0 %, активных сэмплов
3889 -> 2026. Серия E: function в 2x1.0 CPU — потолок 552 (530–552)
успешных/с, упор в CPU API; в 3x1.0 / БД 0.5 — ~420–460, упор в БД.
analyze.py: линии без заливки, цвет = конфигурация, пунктир = multi,
подпись — максимум серии. REPORT: раздел трека 1 и ограничения.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 17:33:46 +03:00
Alexey BodanyukandClaude Opus 5.5 479cd78f7f bench: трек 1 — серия D (A/B multi vs function, 3 повтора), план серии E
Образ коммита 4830be4, одна переменная MESSAGE_WRITE_PATH:
2x1.0 CPU / БД 2.0 — 283 -> 489 успешных/с (function не насыщен на сетке
до 500/с); 3x1.0 / БД 0.5 — 202 -> 452 успешных/с; SQL-операторов на
сообщение 15.4 -> 3.0. Серия E (ceiling) — поиск потолка нового пути.
Профили py-spy после серии не запустились (series.sh правился во время
выполнения), запускаются отдельно.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 17:08:49 +03:00
Alexey BodanyukandClaude Opus 5.5 55210c6e9c docs: DEFENSE и WINDOWS под lab3, модель данных, CI, сбор доказательств
- DEFENSE: что изменилось с lab2, путь POST с ключом, ответы на вопросы
  README, ловушки из чужих PR, периоды проверок и время обнаружения.
- WINDOWS: запуск на ноутбуке, make из PowerShell (нужен sh из Git),
  сценарий защиты по шагам README.
- DATA_MODEL: idempotency_keys и функция post_message.
- CI: test-failover с восстановлением, свои тесты.
- scripts/collect_evidence.py: все проверки на чистом clone -> evidence/.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 16:03:23 +03:00
Alexey BodanyukandClaude Opus 5.5 8964c9b554 bench: трек 1 — базовые серии A/B/C (3 повтора) и анализатор
Замеры образа коммита 3f45eb1 (до улучшения): фиксированный суммарный CPU
API (1x2.0 / 2x1.0 / 4x0.5), рост числа реплик против узкой БД (0.5 CPU),
рост CPU БД при 3 репликах. raw JSON на каждую ступень, results.csv,
лог прогона. analyze.py: медиана и разброс по повторам, потолок и
SLO-ёмкость, узкое место по CPU относительно фактического лимита,
SVG-графики.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 16:03:23 +03:00
Alexey BodanyukandClaude Opus 5.5 4830be414e perf: идемпотентная отправка сообщения одним вызовом SQL (трек 1)
Гипотеза из профиля (py-spy, 2 x 1.0 CPU, 240 req/s): ~57 % CPU реплики
уходит в psycopg — накладные расходы на каждый оператор; на одно
сообщение ~11 обращений к PostgreSQL в двух транзакциях (проверка
членства, BEGIN, ключ, часы чата, сообщение, COMMIT, проверки пула).

- migrations/0003: функция post_message — та же логика и тот же SQL,
  что в многооператорном пути; вызывается в autocommit, поэтому вызов и
  есть транзакция (ключ, часы чата и сообщение — всё или ничего).
- пул: autocommit-соединения; многооператорная работа по-прежнему в явной
  conn.transaction(); ping — один оператор без транзакции.
- MESSAGE_WRITE_PATH=function (по умолчанию) | multi (прежний путь) — A/B
  на одном образе. Порядок ответов одинаков: постороннему всегда 404,
  даже при неверном ключе/теле.
- Оба режима проходят одни и те же тесты идемпотентности и контракта.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 16:03:22 +03:00
Alexey BodanyukandClaude Opus 5.5 26a0d86431 bench: план серий трека 1 (series.sh)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 14:01:56 +03:00
Alexey BodanyukandClaude Opus 5.5 3f45eb1b51 fix: реплика не «замерзает» при потере сети к PostgreSQL; сети edge/data
Найдено тестом сценария рецензента (docker network disconnect одной API
от сети хранилищ): по дедлайну задача psycopg отменялась, psycopg слал
cancel-запрос, а libpq PQcancelStart синхронно резолвит имя хоста в потоке
event loop — вся реплика, включая /health/live, висела ~1 минуту и не
возвращалась в ротацию сама. Теперь задача по дедлайну не отменяется
(запрос сразу получает 503, задача доживает сама), а мёртвые соединения
рвёт TCP keepalive + tcp_user_timeout = DB_TIMEOUT_SECONDS.

- compose: сеть edge (lb, api) и internal-сеть data (api, migrate,
  postgres, valkey) — LB не видит хранилища.
- тесты: неготовая реплика (pg_hba и network disconnect) выводится из
  ротации, процесс отвечает на live, после восстановления сам UP за <= 15 с;
  пробельные Unicode-строки, 1e1000, байт FF в пароле, 101 участник,
  курсор одного списка в другом.
- bench/track1: CPU реплик суммируется по каждому замеру, в raw и CSV —
  фактические лимиты контейнеров из docker inspect.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 14:01:11 +03:00
Alexey BodanyukandClaude Opus 5.5 2033440cdb bench: трек 1 — оркестратор серий, k6 open-loop, py-spy sidecar
- bench/track1/run.py: на каждую конфигурацию x повтор — свежий стенд
  (отдельный compose-проект, пустые volumes), образ API из git-коммита
  (git archive), ступени constant-arrival-rate; CPU каждого контейнера,
  ожидание пула (STATS_LOG_SECONDS), pg_stat_database/pg_stat_statements,
  счётчики HAProxy; raw JSON + results.csv, без токенов.
- compose.bench.yaml: cpuset для k6 / LB+Valkey / PostgreSQL / API, чтобы
  генератор и LB не отнимали CPU у реплик.
- bench/track1/profile.py + pyspy/: py-spy с хешем пакета, профиль одной
  реплики под нагрузкой.
- API: STATS_LOG_SECONDS (по умолчанию выключено) — счётчики пула и CPU
  процесса в лог; API_IMAGE в compose для образа из коммита.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-16 13:51:37 +03:00
Alexey BodanyukandClaude Opus 5.5 c050642ec2 feat: rolling update без простоя и серия отказов реплик под нагрузкой (уровень 4)
- scripts/rolling-update.sh: сборка, однократные миграции новым образом,
  затем для каждой старой реплики: +1 новая (healthy и UP в HAProxy),
  graceful stop старой; ревизия образа видна в label и логе "starting".
- bench/loadgen.py: фоновая нагрузка через LB (POST с Idempotency-Key +
  GET), неопределённый POST повторяется тем же ключом, проверка истории
  на потери и дубли.
- bench/failover_series.py: каждая реплика по очереди в режимах stop /
  kill / pause; RPS, ошибки, время выхода из ротации и возврата.
- bench/rolling_under_load.py: rolling update под нагрузкой.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-16 13:39:54 +03:00
Alexey BodanyukandClaude Opus 5.5 78b95acabf test: строгие тесты lab3 — идемпотентность, балансировка, отказы реплик
- test_idempotency: точные коды и тела повторов через разные реплики, 24
  одновременных запроса с одним ключом -> ровно одно сообщение и одна строка
  ключа в БД, 409 при другом body, scope user+chat+key, 404 постороннему,
  откат после сбоя на втором INSERT (триггер, n_tup_ins), неопределённый
  POST (клиентский timeout) и lock_timeout с повтором тем же ключом.
- test_balancing: X-Instance-Id = реальные контейнеры, нет sticky,
  LB исключает живую, но неготовую реплику (pg_hba reject), graceful stop
  каждой реплики под нагрузкой без единой ошибки, SIGKILL — ограниченные
  ошибки и восстановление, полный restart и scale up/down без рестарта LB,
  политика compose (лимиты, порты, secrets, digest), миграции один раз.
- Входы из замечаний к чужим PR: курсоры с []/{}/числами/пустые/пробел,
  байт FF, TEXT/text, дубли ключей, Bearer é, UUID в разном регистре, тело
  при входе, пароль "я:"*64. Тесты lab2 переведены на Idempotency-Key и LB;
  все отказы восстанавливаются в finally и ждут UP всех реплик в HAProxy.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-16 13:32:13 +03:00
Alexey BodanyukandClaude Opus 5.5 3f12698676 feat: HAProxy и N API-реплик, X-Instance-Id, общая идемпотентность, drain при SIGTERM
- compose: один масштабируемый сервис api (deploy.replicas=2, без ports и
  container_name), HAProxy 3.2 LTS по digest — единственная точка входа
  (127.0.0.1), разовый migrate до старта реплик, CPU/RAM-лимиты у всех сервисов,
  каждому сервису только его secrets.
- HAProxy: server-template + Docker DNS (re-resolve 1 с), активная проверка
  GET /health/ready каждую 1 с (fall 2 / rise 2), без backup-серверов,
  retry только conn-failure, ошибки самого LB — в формате контракта с X-Request-Id.
- API: X-Instance-Id = hostname контейнера на каждом ответе (и на 500/503);
  Idempotency-Key обязателен; ключ, сообщение и результат пишутся в ОДНОЙ
  транзакции (миграция 0002), конкурентный повтор ждёт уникальный индекс;
  другой body — 409, посторонний — всегда 404; attachment_ids=[] равен
  отсутствию поля; дубли ключей JSON и тело у POST /auth/sessions — 400.
- SIGTERM: readiness 503 + DRAIN_SECONDS=3 обслуживания, затем graceful stop.
- scripts: stop-one.sh (+ .cmd для Windows), restore-api.sh, restart.sh
  перезапускает и LB, lbctl.py — состояние бэкендов HAProxy -> контейнеры.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-16 13:32:13 +03:00
Alexey BodanyukandClaude Opus 5.5 4f6cfda37a chore: перенос решения lab2 (src, миграции, student-tests, Dockerfile, compose, lock) без изменений
Исходная точка lab3: код из solution/bvt2302-bodanyuk лабы 2, коммит ef44e48
(PR mtusi2026/mtusi2026lab2#42 принят). README, COURSE, contracts/ и tests/
преподавателя не тронуты; к Makefile дописан студенческий раздел lab2,
.env.example — версия lab2. Исследование трека 2 lab2 (bench/track2) не переносилось.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-16 13:02:13 +03:00
386 changed files with 62289 additions and 18 deletions
+29 -1
View File
@@ -5,9 +5,37 @@ APP_PORT=8080
APP_ENV=lab
SESSION_TTL_SECONDS=3600
# Choose credentials locally. File paths contain no secret values.
# The files are created by `python scripts/init_secrets.py` and mounted by compose as secrets.
DATABASE_URL_FILE=/run/secrets/database_url
# Optional on 3, required use on 4 in lab2; Redis OR Valkey:
SESSION_STORE=postgres
# valkey = sessions live only in Valkey (AOF, appendfsync always); postgres = table `sessions`.
SESSION_STORE=valkey
CACHE_URL_FILE=/run/secrets/cache_url
# Number of interchangeable API replicas behind HAProxy (compose deploy.replicas).
API_REPLICAS=2
# Idempotency-Key results are kept at least this long (contract: >= 24 h).
IDEMPOTENCY_TTL_SECONDS=86400
# Write path of POST message: function = one SQL call (migration 0003), multi = lab3 baseline path.
MESSAGE_WRITE_PATH=function
# Additional settings of this implementation (defaults shown; all optional):
LOG_LEVEL=info
DB_POOL_MIN=1
DB_POOL_MAX=10
DB_TIMEOUT_SECONDS=3
CACHE_TIMEOUT_SECONDS=0.5
PASSWORD_HASH_CONCURRENCY=2
DOCS_ENABLED=true
# Graceful stop of a replica: readiness 503 + drain, then up to SHUTDOWN for in-flight requests.
DRAIN_SECONDS=3
SHUTDOWN_TIMEOUT_SECONDS=8
# CPU/RAM limits (docker compose deploy.resources.limits); API_* are per replica.
API_CPUS=1.0
API_MEMORY=256m
PG_CPUS=2.0
PG_MEMORY=1g
VALKEY_CPUS=0.5
LB_CPUS=1.0
PG_MAX_CONNECTIONS=100
# Host ports of the load balancer (published on 127.0.0.1 only; API replicas publish nothing)
APP_PUBLISH_PORT=8080
LB_STATS_PORT=8404
+5
View File
@@ -0,0 +1,5 @@
# LF everywhere: identical files on Windows/Linux and inside `git archive | docker build -`.
* text=auto eol=lf
*.png binary
# cmd.exe expects CRLF in batch files.
*.cmd text eol=crlf
+36
View File
@@ -0,0 +1,36 @@
name: Application (lab3)
on: [push, pull_request]
permissions:
contents: read
jobs:
app:
runs-on: ubuntu-latest
timeout-minutes: 40
steps:
- uses: actions/checkout@v4
- uses: actions/setup-python@v5
with:
python-version: '3.13'
- name: Synthetic secrets and env (CI only, never committed)
run: python scripts/init_secrets.py && cp .env.example .env
- name: Build, migrate once, start 2 API replicas behind HAProxy, wait for healthy
run: docker compose --env-file .env up -d --build --wait
- name: Contract and public smoke
run: make check test PYTHON=python
- name: Persistence across restart of LB, API replicas and stores
run: make test-persistence ACTION_SCRIPT=scripts/restart.sh PYTHON=python
- name: Failover of the replica that served the test, then restore it
run: |
make test-failover ACTION_SCRIPT=scripts/stop-one.sh PYTHON=python
sh scripts/restore-api.sh && python scripts/lbctl.py wait-up 2
- name: Own tests (unit + stand; every failure is restored in finally)
run: |
python -m pip install --require-hashes --no-deps -r requirements.txt
python -m pip install -r requirements-test.txt
make student-test PYTHON=python
- name: Logs
if: always()
run: docker compose logs --tail 100 api lb
- name: Remove CI stand and its synthetic data
if: always()
run: docker compose down -v
+59
View File
@@ -0,0 +1,59 @@
# syntax=docker/dockerfile:1
# Messenger API, lab 3 (replicas behind HAProxy, PostgreSQL + Valkey). Built by compose.yaml (services api, migrate).
#
# Python is interpreted, so there is no separate compile stage; the image
# contains only the interpreter, locked dependencies and the source code.
# Base image is pinned by version AND digest (python 3.13.16, Debian 12 bookworm).
FROM python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
ENV PYTHONDONTWRITEBYTECODE=1 \
PYTHONUNBUFFERED=1 \
PIP_NO_CACHE_DIR=1 \
PIP_DISABLE_PIP_VERSION_CHECK=1
WORKDIR /app
# 1. Dependencies first (layer cache survives source edits). Every package,
# including transitive ones, is pinned with sha256 hashes in requirements.txt;
# --no-deps guarantees nothing outside the lock file is installed.
COPY requirements.txt ./
RUN pip install --root-user-action=ignore --require-hashes --no-deps -r requirements.txt
# 2. Unprivileged runtime user. Code stays owned by root, so the process
# cannot modify its own code.
RUN groupadd --system --gid 10001 app \
&& useradd --system --uid 10001 --gid app --no-create-home --shell /usr/sbin/nologin app
COPY src/ ./src/
# Versioned SQL migrations, applied by `python -m messenger.migrate` (compose service migrate).
COPY migrations/ ./migrations/
# Course contract, served read-only at /openapi.json for Swagger UI (/docs).
COPY contracts/openapi.json ./contracts/openapi.json
# Non-secret defaults; compose.yaml passes the rest. Passwords come only as files
# (DATABASE_URL_FILE / CACHE_URL_FILE -> Docker secrets), never as ENV or COPY.
ENV PYTHONPATH=/app/src \
APP_HOST=0.0.0.0 \
APP_PORT=8080 \
APP_ENV=lab \
LOG_LEVEL=info \
SHUTDOWN_TIMEOUT_SECONDS=8 \
DOCS_ENABLED=true \
SESSION_TTL_SECONDS=3600 \
DATABASE_URL_FILE=/run/secrets/database_url
# Build revision (scripts/rolling-update.sh passes git commit + time): makes every rolled-out
# image distinguishable in `docker inspect` and in the "starting" log line of each replica.
ARG BUILD_REVISION=dev
LABEL org.opencontainers.image.revision=$BUILD_REVISION
ENV APP_REVISION=$BUILD_REVISION
USER app:app
EXPOSE 8080
STOPSIGNAL SIGTERM
HEALTHCHECK --interval=10s --timeout=3s --start-period=5s --retries=3 \
CMD ["python", "-m", "messenger.healthcheck"]
# Exec form: the Python server is PID 1 and receives SIGTERM directly (no shell in between).
CMD ["python", "-m", "messenger"]
+53
View File
@@ -28,3 +28,56 @@ test-persistence:
test-failover:
@test -n "$(ACTION_SCRIPT)" || (echo "Set ACTION_SCRIPT to your implemented executable stop-one hook"; exit 2)
$(PYTHON) tests/resilience.py --mode failover --action-script "$(ACTION_SCRIPT)" --base-url "$(BASE_URL)" $(TLS_ARGS)
# ---------------------------------------------------------------------------
# Student solution (lab 3, Python). On Windows: PowerShell/Git Bash with GNU make, public
# tests need PYTHONUTF8=1, and the hooks are started through scripts/*.cmd (Git Bash).
.PHONY: secrets up down logs ps migrate student-test unit-test explain lb-status restore rolling-update failover-series
secrets:
$(PYTHON) scripts/init_secrets.py
# Build, apply migrations ONCE (service migrate), start API replicas and the LB, wait for healthy.
up: secrets
docker compose up -d --build --wait
# Safe stop: containers stop, named volumes (data) stay.
down:
docker compose down
logs:
docker compose logs --tail 50 api lb
ps:
docker compose ps -a
migrate:
docker compose run --rm migrate
# HAProxy view of the replicas (UP/DOWN) mapped to containers and X-Instance-Id.
lb-status:
$(PYTHON) scripts/lbctl.py status
$(PYTHON) scripts/lbctl.py replicas
# Start replicas stopped by scripts/stop-one.sh again and wait until the LB has them all UP.
restore:
sh scripts/restore-api.sh
$(PYTHON) scripts/lbctl.py wait-up $${API_REPLICAS:-2}
# Replace every API replica one by one with a freshly built image (no downtime).
rolling-update:
sh scripts/rolling-update.sh
# Level 4: stop each replica in turn under load, measure RPS/errors/recovery (evidence/failover/).
failover-series:
$(PYTHON) bench/failover_series.py --output evidence/failover
# Own tests: unit + black-box against the running stand (needs requirements*.txt in a venv).
student-test:
$(PYTHON) -m pytest student-tests -v
unit-test:
$(PYTHON) -m pytest student-tests -m "not stack" -q
explain:
$(PYTHON) scripts/explain_history.py --output evidence/explain-history.txt
+268 -17
View File
@@ -1,37 +1,288 @@
# Отчёт — лабораторная 3
- Автор, группа: TODO
- Целевая оценка; выбранный трек 5, если нужен: TODO
- Commit/PR: TODO после публикации
- Использованные источники, библиотеки и помощь ИИ: TODO
- Автор, группа: Боданюк Алексей, БВТ2302
- Целевая оценка: **5**, трек 1 «Цена масштабирования»
- Ветка: `solution/bvt2302-bodanyuk`. Форк: `https://gitea.privcloud.tech/wfa/mtusi2026lab3`. PR: «ЛР 3 — БВТ2302 — Боданюк — на 5» в mtusi2026/mtusi2026lab3
- Стек: Python 3.13.16, Starlette 1.7.0, uvicorn 0.54.0, psycopg 3.3.6 + psycopg-pool 3.3.3, redis-py 8.1.0, argon2-cffi 25.1.0; PostgreSQL 18.6, Valkey 8.1.3, HAProxy 3.2.26 (LTS); замеры — k6 2.3.0, py-spy 0.4.2. Все образы закреплены по тегу **и** digest, все Python-пакеты — по sha256 ([requirements.txt](requirements.txt), [bench/track1/pyspy/requirements.txt](bench/track1/pyspy/requirements.txt))
- Источники: README, COURSE, контракт ([openapi.json](contracts/openapi.json), [семантика](contracts/README.md)), документация HAProxy 3.2 (server-template, resolvers, http-check, retry-on, http-error), Docker Compose (deploy, networks), PostgreSQL (INSERT … ON CONFLICT, PL/pgSQL, pg_stat_statements, libpq keepalives/tcp_user_timeout), psycopg 3 (pool, autocommit), k6 (constant-arrival-rate, dropped_iterations). Учтены все замечания преподавателя к PR lab3 и lab2 других студентов ([раздел ниже](#проверки-на-которых-ловили-другие-работы)); мой PR lab2 (#42) принят без замечаний
- Помощь ИИ: код, тесты, скрипты экспериментов и черновик отчёта написаны с помощью ИИ-ассистента Claude Code (Anthropic). Все команды и замеры реально выполнены на стенде ниже; вывод — в [evidence/](evidence/). Ответы на вопросы защиты — [docs/DEFENSE.md](docs/DEFENSE.md), запуск на Windows и сценарий защиты по шагам — [docs/WINDOWS.md](docs/WINDOWS.md)
## Запуск из чистого clone
TODO: версии runtime/образов, prerequisites, команды создания локальных secrets/.env, build, миграций, запуска, ожидания готовности, тестов и остановки. Указать необходимые действия оператора. Не вставлять секреты. Отдельно указать безопасную остановку и команду намеренного удаления учебных данных.
Нужны Git, Docker с Compose v2 (проверено: Docker 29.8.0, Compose 5.5.1), Python 3.9+ и GNU make (на Windows — см. [WINDOWS.md](docs/WINDOWS.md)). Учётные данные в Git не хранятся, их создаёт скрипт.
```sh
git clone https://gitea.privcloud.tech/wfa/mtusi2026lab3 mtusi2026lab3 && cd mtusi2026lab3 && git checkout solution/bvt2302-bodanyuk
python3 scripts/init_secrets.py # случайные локальные пароли -> ./secrets (git-ignored), идемпотентно
cp .env.example .env # API_REPLICAS=2, лимиты, таймауты; значения по умолчанию есть и в compose.yaml
docker compose --env-file .env config --quiet
docker compose --env-file .env up -d --build --wait
# postgres, valkey healthy -> migrate (один раз, exit 0) -> api x2 healthy -> lb healthy
curl -i http://127.0.0.1:8080/health/ready # 200, X-Instance-Id: <container id>; Swagger /docs; LB: http://127.0.0.1:8404/
make check && make test
make test-persistence ACTION_SCRIPT=scripts/restart.sh
make test-failover ACTION_SCRIPT=scripts/stop-one.sh && make restore
```
- **Windows:** публичным тестам нужен `PYTHONUTF8=1`; Windows не исполняет `.sh`, поэтому хуки запускаются через обёртки `scripts/restart.cmd` и `scripts/stop-one.cmd` (тот же `.sh` в Git Bash); рецептам Makefile нужен `sh` из `C:\Program Files\Git\usr\bin`.
- **Свои тесты:** `python3 -m venv .venv`, `pip install --require-hashes --no-deps -r requirements.txt`, `pip install -r requirements-test.txt`, затем `make student-test` (~8 мин, стенд должен быть поднят). Без стенда — `make unit-test`.
- **Все доказательства одной командой** (на чистом clone, ~25 мин): `python scripts/collect_evidence.py`.
- **Миграции:** только сервис `migrate` (`python -m messenger.migrate`, advisory lock, checksum файлов), до старта реплик. Повтор: `make migrate` → `already applied, skipped`.
- **Остановка:** безопасно — `docker compose down` (volumes сохраняются). Намеренное удаление учебных данных — `docker compose down -v` и `rm -rf secrets`.
## Архитектура и решения
TODO: схема процессов/хранилищ, источник истины, транзакционные границы, ограничения. Что перенесено из lab2, что изменено в этой работе. Соответствие переменных из .env.example реальной конфигурации.
```text
клиент ──127.0.0.1:8080──> lb: HAProxy 3.2 ──сеть edge──> api × N (Starlette/uvicorn, без состояния, X-Instance-Id = hostname)
│ check GET /health/ready 1 с, fall 2/rise 2 │
│ Docker DNS `api` каждые 1 с (server-template, 8 слотов) └──сеть data (internal)──> postgres 18.6 (источник истины)
└ свои 502/503/504 — JSON по контракту с X-Request-Id valkey 8.1.3 (сессии, AOF always)
migrate (одноразовый) ──data──> postgres: миграции 0001–0003 до старта реплик
```
| Роль | Service | Образ | Health / readiness | Volume | Сеть / порт | Credentials | CPU / RAM |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Точка входа | `lb` | `haproxy:3.2.26-alpine@sha256:254e6370…` | `haproxy -c` + stats; бэкенды — активная проверка `/health/ready` | — | edge; `127.0.0.1:8080` и `[::1]:8080` (loopback, см. ниже), `127.0.0.1:8404` (stats) | нет | 1.0 / 128m |
| API × 2 | `api` (`deploy.replicas`) | `mtusi-messenger:lab3` из исходников | Docker: `/health/ready` 2 с; LB: своя проверка | — | edge + data, только `expose 8080` | secrets `database_url`, `cache_url` | 1.0 / 256m каждая |
| Миграции | `migrate` | тот же образ | exit 0 | — | data | secret `database_url` | 0.5 / 256m |
| БД | `postgres` | `postgres:18.6-bookworm@sha256:afc7e2d4…` | `pg_isready` | `pgdata` | data (internal) | secret `postgres_password` | 2.0 / 1g |
| Сессии | `valkey` | `valkey/valkey:8.1.3-bookworm@sha256:fea8b3e6…` | `valkey-cli ping` | `valkeydata` | data (internal) | secret `valkey_password` → include-файл | 0.5 / 256m |
**Что перенесено из lab2** (отдельный коммит `4f6cfda`, без изменений): API, PostgreSQL-схема 0001, сессии только в Valkey, Argon2id, keyset-cursor, строгая валидация, дедлайны БД/Valkey, собственные тесты. **Что изменено в lab3:**
- **Балансировка без sticky.** Один Compose-сервис `api` масштабируется (`--scale`, `deploy.replicas`), без `container_name` и `ports`. HAProxy находит реплики через Docker DNS (`server-template api 8 api:8080 … resolvers docker`, `hold valid 1s`), поэтому scale up/down, новый IP после рестарта и полный restart стенда подхватываются без перезапуска LB (проверено тестами).
- **Readiness управляет маршрутизацией.** `option httpchk GET /health/ready`, `inter 1s fastinter 500ms fall 2 rise 2`, `timeout check 1s`, без backup-слотов. Ready = PostgreSQL и Valkey отвечают и нет drain. Реплика, живая по `/health/live`, но без БД/Valkey, не получает ни одного бизнес-запроса.
- **Повторы в LB только безопасные.** `retry-on conn-failure` + `option redispatch`: повторяется только запрос, который не удалось отправить реплике. Всё, что могло дойти до API (включая неидемпотентные POST создания чата/пользователя), повторяет клиент сам; отправку сообщения — с тем же `Idempotency-Key`.
- **Graceful shutdown.** SIGTERM → readiness 503 сразу → `DRAIN_SECONDS=3` реплика ещё обслуживает (LB успевает пометить её DOWN) → uvicorn закрывает сокет, запросы в полёте ≤ `SHUTDOWN_TIMEOUT_SECONDS=8` → exit 0. 3 + 8 < `stop_grace_period: 15s`, SIGKILL не нужен.
- **`X-Instance-Id`** = hostname контейнера (короткий container ID): стабилен для процесса, однозначно сопоставляется контейнеру (`docker inspect`), что использует `scripts/stop-one.sh`. Middleware ставит его и `X-Request-Id` на **каждый** ответ API, включая 404/405/413/500/503. Ошибки, созданные самим HAProxy, — без `X-Instance-Id` (их создал не API), но в формате контракта и с тем же `X-Request-Id`.
- **Общая идемпотентность** ([0002](migrations/0002_idempotency_keys.sql), [0003](migrations/0003_post_message_function.sql), [repo.py](src/messenger/repo.py), [app.py](src/messenger/app.py)). Таблица `idempotency_keys (user_id, chat_id, key)` → SHA-256 канонического тела → `message_id`. Ключ, часы чата, сообщение и «результат» — **одна транзакция**; отдельного pending-состояния нет, поэтому сбой не оставляет «висящий» ключ, а транзиентная 5xx не фиксируется. Конкурентный повтор ждёт на уникальном индексе завершения первой транзакции и затем видит её результат (READ COMMITTED). Тот же ключ и тело → тот же 201 и Message; другое тело → 409; постороннему — 404 всегда (членство проверяется до ключа); `attachment_ids: []` = отсутствие поля, порядок ключей JSON не важен. Ключи хранятся ≥ 24 ч, старше — очищаются пачками раз в 10 мин под advisory-замком.
- **Миграции** — только сервис `migrate`, до старта реплик; 0002 и 0003 только добавляют объекты и совместимы с работающей предыдущей версией (rolling update: сначала миграция новым образом, затем замена реплик). При `docker compose start api` (например, в `restart.sh`) Compose снова запускает `migrate` как зависимость — он берёт advisory-замок и отвечает `already applied, skipped`; реплики ждут его завершения ([test-persistence.txt](evidence/test-persistence.txt)).
- **Порты только на loopback.** LB опубликован на `127.0.0.1` и `[::1]`. На Windows `localhost` сначала разрешается в `::1`, и отказ соединения по нему стоил ~2 с на **каждый** запрос (измерено: 2,0 с против 0,007 с). Публичные тесты по умолчанию ходят на `http://localhost:8080`, и в первом прогоне из чистого clone `make test-failover` из-за этого успел сделать лишь 8 записей за 15 с. После исправления — 0,013 с.
- **Сети.** `edge` (lb, api) и internal `data` (api, migrate, postgres, valkey): LB не видит хранилища, у хранилищ нет выхода наружу. Secrets — каждому сервису только свои (проверяется тестом по `docker compose config`).
- **Дефект, найденный на lab3-сценарии рецензента.** Если одной реплике отключить сеть `data`, по дедлайну задача psycopg отменялась, psycopg слал cancel-запрос, а libpq `PQcancelStart` синхронно резолвит имя хоста **в потоке event loop** — вся реплика, включая `/health/live`, зависала примерно на минуту и не возвращалась сама. Теперь по дедлайну запрос сразу получает 503, а задача не отменяется и завершается сама; мёртвые соединения рвёт TCP keepalive + `tcp_user_timeout` = 3 с ([db.py](src/messenger/db.py)). Повторная проверка: live отвечает за 0,00 с, ready 503 за 1,0 с, после восстановления сети реплика сама UP за ~2 с (тест `test_lb_stops_routing_…[network]`).
- **Улучшение трека 1** — запись сообщения одним вызовом SQL-функции `post_message` (подробно в треке); прежний путь доступен как `MESSAGE_WRITE_PATH=multi`, оба проходят одни и те же тесты.
Readiness-критичные зависимости: PostgreSQL (всё, кроме health) и Valkey (все защищённые запросы). Swagger/stats LB на readiness не влияют.
Соответствие `.env.example` реальной конфигурации ([config.py](src/messenger/config.py), [compose.yaml](compose.yaml)):
| Переменная | Значение | Где используется |
| --- | --- | --- |
| `APP_HOST`, `APP_PORT`, `APP_ENV` | `0.0.0.0`, `8080`, `lab` | bind uvicorn, стартовый лог |
| `SESSION_TTL_SECONDS` | `3600` | срок сессии (compose передаёт в каждую реплику) |
| `DATABASE_URL_FILE`, `CACHE_URL_FILE` | `/run/secrets/…` | строки подключения из secret-файлов |
| `SESSION_STORE` | `valkey` (или `postgres`) | хранилище сессий |
| `API_REPLICAS` | `2` | `deploy.replicas` сервиса `api` |
| `IDEMPOTENCY_TTL_SECONDS` | `86400` (меньше нельзя) | срок хранения результата ключа |
| `MESSAGE_WRITE_PATH` | `function` (или `multi`) | путь записи сообщения (трек 1, A/B) |
| `DRAIN_SECONDS`, `SHUTDOWN_TIMEOUT_SECONDS` | `3`, `8` | graceful shutdown |
| `DB_POOL_MIN/MAX`, `DB_TIMEOUT_SECONDS`, `CACHE_TIMEOUT_SECONDS`, `PASSWORD_HASH_CONCURRENCY` | 1/10, 3, 0.5, 2 | пул, дедлайны, параллельность Argon2id |
| `API_CPUS/MEMORY`, `PG_CPUS/MEMORY`, `VALKEY_CPUS`, `LB_CPUS`, `PG_MAX_CONNECTIONS` | 1.0/256m, 2.0/1g, 0.5, 1.0, 100 | лимиты контейнеров |
| `APP_PUBLISH_PORT`, `LB_STATS_PORT` | `8080`, `8404` | порты LB на `127.0.0.1` (API портов не публикует) |
| `LOG_LEVEL`, `DOCS_ENABLED` | `info`, `true` | логи, Swagger |
## Доказательства критериев
| Критерий README | Команда/сценарий | Наблюдаемый результат | Файл доказательства |
Стенд: Windows 10 22H2, AMD Ryzen 5 3400G (4 ядра / 8 потоков), 30 ГБ RAM, Docker Desktop 29.8.0 (WSL2, 8 vCPU), Python 3.14 на хосте, GNU make 4.4.1. Все проверки уровней 3 и 4 ниже выполнены 2026-10-19 **на стенде из чистого clone** (новая папка, `git clone`, пустые volumes, secrets созданы заново) одной командой `python scripts/collect_evidence.py`, коммит `2708732`. Вывод сохранён целиком.
| Критерий README | Команда / сценарий | Наблюдаемый результат | Файл |
| --- | --- | --- | --- |
| Публичный smoke | make test | TODO | TODO |
| Остальные пункты уровня 3 | TODO | TODO | TODO |
| Все пункты уровня 4, если заявлен | TODO | TODO | TODO |
| Выбранный трек 5, если заявлен | TODO | TODO | TODO |
| Чистый clone, secrets, миграции с нуля, одна команда запуска | `init_secrets.py`, `compose config --quiet`, `compose up -d --build --wait`, `compose logs migrate`, `git check-ignore` | все 5 сервисов healthy, `migrate` Exited (0): `0001…0003: applied`; образы по digest; `secrets/` и `.env` игнорируются Git | [clean-clone.txt](evidence/clean-clone.txt) |
| Контракт | `make check`, `make validate` | PASS, PASS | [make-check-validate-test.txt](evidence/make-check-validate-test.txt) |
| **3:** публичный smoke | `make test` | `PASS: lab3 public API scenario` | там же |
| **3:** ≥ 2 реплики, одна точка входа, общая БД и сессии | `lbctl.py replicas/status`, `test_one_session_is_served_by_every_real_replica`, `test_write_on_one_replica_is_read_on_the_other` | 2 контейнера `api` UP в HAProxy; один токен — ответы от обоих, `X-Instance-Id` = их hostname; запись на одной сразу читается на другой; logout сразу действует на обеих | [clean-clone.txt](evidence/clean-clone.txt), [student-tests.txt](evidence/student-tests.txt) |
| **3:** остановка любой реплики, ≤ 10 с, подтверждённое на месте | `make test-failover ACTION_SCRIPT=scripts/stop-one.cmd` → `make restore` → снова (по очереди обе) | PASS для `api-1` и для `api-2`: 131 / 132 успешных записи, **0 неудачных попыток**, запись после хука — 5,9 / 5,7 с; остановленный контейнер `exited (0)`, вторая реплика UP | [test-failover.txt](evidence/test-failover.txt) |
| **3:** повтор POST без дубля, в т.ч. конкурентный и через другую реплику; неподтверждённую запись можно повторить | `test_idempotency.py` (21 тест) + публичный failover-тест (повторяет неопределённый POST с тем же ключом) | см. таблицу «ловушек» ниже; публичный тест: дублей нет | [student-tests.txt](evidence/student-tests.txt) |
| **3:** persistence | `make test-persistence ACTION_SCRIPT=scripts/restart.cmd` | PASS; у `lb`, `api-1`, `api-2`, `postgres`, `valkey` новые PID и `StartedAt`, volumes те же, токен тот же | [test-persistence.txt](evidence/test-persistence.txt) |
| **4:** readiness, таймауты, SIGTERM согласованы; LB исключает недоступный backend | `test_lb_stops_routing_…[pg_hba/network × 2]`, `test_graceful_stop_…[×2]`, `test_graceful_shutdown_order_in_logs` | живая, но неготовая реплика выходит из ротации за 3–4 с, live 200 / ready 503, сама возвращается ≤ 15 с; graceful stop под нагрузкой — 0 ошибок; в логах `drain_started` → `shutdown_requested` → `stopped`, exit 0, в логе HAProxy — `is DOWN` для этой реплики | [student-tests.txt](evidence/student-tests.txt) |
| **4:** период проверок и максимальное время обнаружения | [haproxy.cfg](infra/haproxy/haproxy.cfg) + серия отказов | проверка раз в 1 с, `fall 2`/`rise 2`, `timeout check 1s`; максимум ≈ 3,5 с (заморозка); подробно — в следующем разделе | [failover/series.md](evidence/failover/series.md) |
| **4:** поочерёдная остановка каждой реплики: RPS, ошибки, восстановление | `bench/failover_series.py` (stop / kill / pause × обе реплики) | stop: 0 ошибок; kill: 3 оборванных запроса за ≤ 0,25 с; pause: 4 неудачные попытки за ≤ 3,2 с; потерь и дублей нет | [failover/series.md](evidence/failover/series.md), [series.json](evidence/failover/series.json) |
| **4:** общая идемпотентность параллельными запросами | `test_concurrent_same_key_through_different_replicas_creates_exactly_one`, `…_different_bodies_one_wins` | 24 одновременных → 24 × 201, одно тело, ≥ 2 реплики, 1 строка ключа и 1 сообщение (×3 раунда); разные тела → один победитель, остальные 409 | [student-tests.txt](evidence/student-tests.txt) |
| **4:** лимиты CPU/RAM, нет портов в обход LB, secrets по месту, нет `container_name` | `test_compose_policy_limits_ports_secrets_images`, `test_running_containers_have_limits_and_no_bypass_ports` | лимиты у всех 5 сервисов (config и `docker inspect`); порты только у `lb` и только loopback; secrets: lb — нет, migrate — `database_url`, api — `database_url`+`cache_url`; сети edge/data | там же |
| **4:** rolling update под нагрузкой | `bench/rolling_under_load.py` → `scripts/rolling-update.sh` | 57 с, 11 848 запросов, **0 ошибок**; `api-1/2 revision=dev` → `api-3/4 revision=2708732-…`; 7119 подтверждённых сообщений, 0 потерь и дублей | [rolling-update/](evidence/rolling-update/) |
| **4:** миграции однократно и совместимы | `test_migrations_ran_once_by_the_migrate_service`, `test_migrations_rerun_is_a_noop` | применяет только `migrate` (в логах API миграций нет); повтор — `already applied, skipped`; 0002/0003 только добавляют объекты и применяются до замены реплик | [student-tests.txt](evidence/student-tests.txt), [rolling-update/script-output.txt](evidence/rolling-update/script-output.txt) |
| Публичный генератор нагрузки | `tests/load.py --duration 30 --concurrency 4` | 5247 успешных, 0 ошибок, 174 req/s (closed-loop через порт Windows), p50/p95/p99 = 19 / 50 / 65 мс | [load.json](evidence/load.json) |
| Свои тесты | `python -m pytest student-tests -v` | **201 passed** (69 unit + 132 против стенда), 8 мин; все отказы восстанавливаются в `finally` | [student-tests.txt](evidence/student-tests.txt) |
| Оба пути записи эквивалентны | тесты идемпотентности, контракта и БД с `MESSAGE_WRITE_PATH=multi` | 105 passed (режим `function` — весь набор выше) | раздел трека 1 |
| **5:** трек 1 | `bench/track1/series.sh baseline|improved|ceiling`, `profile.py`, `analyze.py` | см. раздел «Трек 1» | [evidence/track1/](evidence/track1/) |
TODO: добавить отдельную строку для каждого заявленного критерия. PASS без команды и наблюдения не является доказательством. Проверка инфраструктурного hook требует подтверждения, какие реальные процессы/хранилища перезапущены.
Первый прогон из чистого clone (тот же сценарий, коммит `2846cae`) нашёл проблему: через `localhost` каждое соединение на Windows стоило 2 с (LB был опубликован только на `127.0.0.1`), и `make test-failover` успел сделать 8 записей вместо ~130. Исправлено публикацией и на `[::1]` (коммит `2708732`). Таблица выше — второй, полный прогон.
## Эксперименты и отказы
## Проверки, на которых ловили другие работы
TODO: гипотеза, hardware/лимиты, версии, объём данных, профиль нагрузки, concurrency/arrival rate, длительность/прогрев, повторы, успешные RPS, ошибки, p50/p95/p99, CPU/RAM. Сохранить raw summary без tokens и объяснить ограничения измерения.
Замечания преподавателя к PR lab3 (#1–#19) и lab2 других студентов, проверенные на своём стенде:
| Воздействие | Что ожидаем | Что наблюдали | Время восстановления / потеря данных |
| Замечание к чужим PR | У нас | Где проверено |
| --- | --- | --- |
| после сбоя ключ остаётся «pending»: повтор висит / 504 / 409 | ключ и сообщение в одной транзакции; сбой триггером **после** вставки ключа → нет ни ключа, ни сообщения (`n_tup_ins` доказывает, что INSERT был), повтор → 201 | `test_failure_after_the_key_insert_rolls_everything_back` |
| одновременные повторы дают 19×201 + 1×409 | 24 одновременных запроса через ≥ 2 реплики, три раунда → 24×201, одно тело, 1 сообщение, 1 строка ключа | `test_concurrent_same_key_through_different_replicas_creates_exactly_one` |
| посторонний: 404, затем 409 | всегда 404, и с чужим ключом, и с другим телом | `test_outsider_and_revoked_session_never_see_the_stored_result` |
| `attachment_ids: []` при повторе → 400 | равно отсутствию поля, тот же 201; порядок ключей не важен | `test_key_order_and_missing_attachment_ids_are_the_same_body` |
| `X-Instance-Id: unknown`; нет заголовков на 500 / 413 | hostname контейнера; middleware на всех ответах, включая 500 из триггера и 413 | `Api.call` проверяет **каждый** ответ; `test_every_api_response_carries_…` |
| LB шлёт запросы неготовой реплике (pg отнят, `network disconnect`, забит пул); fallback-слоты без проверки | активная проверка `/health/ready`, без backup; реплика выходит из ротации за 3,2–4,1 с и сама возвращается | `test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready[pg_hba/network × 2 реплики]` |
| после общего restart LB видит одну реплику | re-resolve Docker DNS каждую секунду | `test_full_restart_lb_rediscovers_all_replicas` |
| «обновление» останавливает обе API; нет проверки под нагрузкой | +1 новая реплика, затем graceful stop старой; под нагрузкой 0 ошибок | [evidence/rolling-update/](evidence/rolling-update/) |
| миграции в каждой API | только `migrate`; в логах API нет миграций | `test_migrations_ran_once_by_the_migrate_service` |
| нет лимитов у LB, migrate, PostgreSQL, Redis | лимиты у всех 5 сервисов; проверены и `compose config`, и `docker inspect` запущенных | `test_compose_policy_…`, `test_running_containers_have_limits_…` |
| неидемпотентные POST повторяются LB | `retry-on conn-failure` — только неотправленные запросы | [haproxy.cfg](infra/haproxy/haproxy.cfg), DEFENSE.md |
| cursor с числом/`null`/`[]`/`{}`/`false`, пустой, пробел → 500 или 200 | 400 на обоих списках; курсор одного списка в другом → 400 | `test_malformed_cursor_values_are_400_on_both_lists`, `test_cursor_of_one_list_is_rejected_by_the_other` |
| `{"text":"first","TEXT":"second"}`, дубли ключей, `{"text":123}`, хвост после JSON, байт FF | 400, ничего не сохраняется | `test_bad_message_body_is_400_and_nothing_is_stored` |
| `Bearer é` → 500; UUID в разном регистре → 503; тело при входе игнорируется | 401; 400 (дубль после нормализации); 400 | `test_non_ascii_or_garbage_bearer_…`, `test_member_ids_are_compared_as_uuids_…`, `test_login_with_a_body_is_rejected` |
| пароль `"a"` принят, `"я:"×64` отклонён; пробельные Unicode-строки (U+00A0, U+2003) приняты; `1e1000` роняет API; 101 участник | 400 / 201; 400; 400 и API живо; 400 | `test_password_range_…`, `test_unicode_whitespace_only_strings_are_400`, `test_odd_registration_bodies_…`, `test_more_than_100_members_is_400` |
| текст обрезается по краям, длина считается после обрезки | хранится как есть; 4001 code point → 400 | `test_text_is_stored_verbatim_and_length_counts_raw_text` |
| трек 1: в «RPS» входят ошибки; неверное суммирование CPU реплик; бюджет в отчёте не совпадает с данными; на графике несопоставимые точки | считаются только 201 (`ok_messages`), dropped отдельно; CPU реплик суммируется в каждый момент; фактические лимиты из `docker inspect` в каждом raw-файле; на графике серии одной конфигурации ресурсов | [analyze.py](bench/track1/analyze.py), [run.py](bench/track1/run.py) |
| трек 1: улучшение не воспроизвелось на другом хосте | улучшение структурное (число обращений к БД на запрос, видно в `pg_stat_statements`), A/B на одном образе одной переменной | серия D |
| lab2: отозванный токен «оживает» после восстановления Redis; БД упала → 401/500 | сессии только в Valkey (нет второй копии); 503 | `test_failures.py`, `test_sessions.py` (перенесены из lab2) |
Сам Content-Type `text/plain` даёт 400, а не 415: в OpenAPI lab3 для `POST /users` и `POST …/messages` ответ 415 не объявлен, контракт требует отклонения — оно есть.
## Эксперименты и отказы (уровень 4)
**Поочерёдная остановка каждой реплики** ([bench/failover_series.py](bench/failover_series.py), [series.md](evidence/failover/series.md)). Нагрузка — 4 closed-loop клиента с хоста через LB: POST сообщения (неопределённый результат → повтор тем же ключом) + GET последней страницы, таймаут клиента 3 с. Раунд: 6 с прогрева → остановка → 12 с без реплики → старт → ожидание UP в HAProxy → 6 с. Перед каждым раундом предыдущая реплика уже восстановлена.
| Воздействие | Что ожидаем | Что наблюдали (api-1 / api-2) | Восстановление / потери |
| --- | --- | --- | --- |
| TODO | TODO | TODO | TODO |
| `docker stop` (SIGTERM, drain) | 0 ошибок: реплика сама уходит из ротации до закрытия сокета | 3568 / 3565 запросов во время отказа, **0 ошибок**; цель последний раз ответила через 1,28 / 1,24 с после SIGTERM (drain) | возврат: UP через 3,3 / 1,4 с, снова обслуживает через 3,4 / 3,1 с; 0 потерь, 0 дублей |
| `docker stop -t 0` (SIGKILL) | обрываются только запросы «в полёте» на убитой реплике | 3 неудачных попытки (обрыв соединения) за ≤ 0,25 с; новые запросы сразу идут на живую (`retry-on conn-failure`) | возврат 2,8 / 3,6 с; неопределённые POST повторены тем же ключом, 0 дублей |
| `docker pause` (заморожен процесс — худший случай) | запросы на замороженную реплику ждут до таймаута, пока проверки её не выключат | 4 неудачные попытки, последняя через 3,2 / 3,1 с; ok-RPS во время отказа падает ~вдвое (часть клиентов ждёт) | после `unpause` UP через 2,2 / 1,2 с; 0 потерь, 0 дублей |
| неготовая, но живая реплика (`pg_hba reject` / `network disconnect` сети `data`) | LB исключает её по readiness, бизнес-запросы к ней не идут | DOWN через 3,2–4,1 с, затем 40 из 40 чтений и 20 из 20 регистраций — через другую реплику; live 200 / ready 503, процесс отвечает | сама UP ≤ 15 с после восстановления доступа (тест) |
| rolling update | ≥ 2 реплики в каждый момент | 57 с, 11 848 запросов, 0 ошибок, p95 53 мс | обе реплики на новой ревизии; 0 потерь, 0 дублей |
| restart всего стенда (`restart.sh`) | данные и токен на месте, LB находит новые IP сам | PASS публичного теста; новые PID у всех 5 процессов | LB видит обе реплики без своего рестарта (тест) |
**Период проверок и максимальное время обнаружения.** HAProxy проверяет `GET /health/ready` каждую 1 с, после первой неудачи — каждые 0,5 с (`fastinter`), DOWN после 2 неудач подряд, UP после 2 удач, таймаут проверки 1 с.
| Ситуация | Максимум до DOWN |
| --- | --- |
| реплика сразу отвечает 503 (drain после SIGTERM) | ≈ 1 + 0,5 = 1,5 с |
| ready отвечает медленно (отнят PostgreSQL/Valkey: ответ ≈ 1 с = таймаут проверки) | ≈ 1 + 1 + 0,5 + 1 = 3,5 с |
| процесс заморожен (каждая проверка ждёт таймаут) | ≈ 1 + 1 + 0,5 + 1 = 3,5 с |
| процесс убит или порт закрыт | новые запросы получают отказ соединения, HAProxy сразу отправляет их на другую реплику (`retry-on conn-failure` + `redispatch`); проверка лишь фиксирует это |
Измерения выше в эти пределы укладываются: при drain реплика выходит из ротации за ≈ 1,2–1,3 с, а 3,2–4,1 с (отнятая БД/сеть, заморозка) включают 0,1–0,5 с опроса через `docker exec`.
**Ограничение измерения.** Нагрузка closed-loop с хоста Windows через проброс порта Docker Desktop (~200–290 ok-RPS). Это проверка доступности и корректности, а не ёмкости — ёмкость меряет трек 1 с k6 внутри сети стенда. Неудачной попыткой считается любой не-2xx ответ или ошибка соединения/таймаут клиента.
## Трек 1. Цена масштабирования
Все цифры ниже — из [evidence/track1/summary.md](evidence/track1/summary.md) (генерирует [analyze.py](bench/track1/analyze.py) из `results.csv`); сырые данные каждой ступени — `evidence/track1/<серия>/<метка>/raw/*.json` (k6 summary, CPU по контейнерам, ожидание пула, счётчики PostgreSQL и HAProxy, фактические лимиты из `docker inspect`). Полный план повторяется одной командой: [bench/track1/series.sh](bench/track1/series.sh).
### Гипотезы
1. Один процесс Python (uvicorn, 1 воркер) исполняет байткод на одном ядре (GIL). Поэтому при фиксированном суммарном CPU одна «толстая» реплика упрётся в одно ядро, а несколько «тонких» используют весь бюджет. Дробить сильнее, чем по ядру на процесс, бесполезно.
2. Реплики API без состояния масштабируют пропускную способность линейно, пока не насытится общий ресурс — PostgreSQL (одна БД на все реплики). После этого добавление реплик не помогает, и потолок задаёт CPU БД. Если дать БД больше CPU, потолок сдвинется — это проверка причины, а не совпадения.
3. LB (HAProxy) и пул соединений не являются узким местом на этих нагрузках: HAProxy тратит единицы процентов ядра на сотни запросов/с.
4. Улучшение: основная цена запроса — накладные расходы на каждый SQL-оператор, а не сама работа БД (гипотеза из профиля). Если выполнить идемпотентную запись одним вызовом, сократятся CPU на сообщение и в API, и в PostgreSQL, и потолок вырастет в обоих режимах — и когда упираемся в API, и когда в БД.
### Методика
- **Стенд.** Ноутбук: AMD Ryzen 5 3400G (4 ядра / 8 потоков SMT), 30 ГБ RAM, Windows 10 22H2, Docker Desktop 29.8.0 (WSL2, ядро 6.18, 8 vCPU, 15,6 ГБ). «1 CPU» в лимитах = 1 vCPU = один аппаратный поток SMT, то есть примерно половина физического ядра.
- **Изоляция.** На время серий основной стенд остановлен. Каждая конфигурация в каждом повторе — **свежий стенд** в отдельном compose-проекте `mtusi-lab3-bench` с пустыми volumes (`down -v` → `up`), чтобы размер истории не копился между конфигурациями. Роли закреплены за своими CPU (`cpuset`, [compose.bench.yaml](bench/track1/compose.bench.yaml)): k6 — 0–1, LB + Valkey — 2, PostgreSQL — 3–4, все реплики API — 5–7. Поверх набора CPU действуют квоты `cpus` из конфигурации (API_CPUS на реплику, PG_CPUS). Генератор и LB не отнимают CPU у API.
- **Код.** Образ API собирается из **коммита** (`git archive <sha> | docker build`), а не из рабочей копии: базовые серии A–C — коммит `3f45eb1` (до улучшения), серии D–E — `4830be4` (с улучшением, старый путь доступен через `MESSAGE_WRITE_PATH=multi`). Фактический образ и лимиты записаны в каждом raw-файле.
- **Нагрузка.** k6 2.3.0 в отдельном контейнере в сети `edge` идёт через LB, как внешний клиент, без проброса портов Windows. Open-loop `constant-arrival-rate`: итерации стартуют с заданной частотой независимо от ответов, поэтому нет coordinated omission. Итерации, которым не хватило VU, считаются в `dropped_iterations` как неудовлетворённый спрос. Одна итерация = `POST /chats/{id}/messages` с новым `Idempotency-Key` в один из 64 чатов (64 пользователя, у каждого свой чат), чтобы блокировка строки одного чата не доминировала. Пользователи и сессии создаются оркестратором до старта k6, файл с синтетическими токенами монтируется только для чтения и удаляется после серии.
- **Ступень:** прогрев 10 с (не учитывается) + замер 20 с. Ступени по возрастанию предложенной нагрузки, 3 повтора. Таймаут запроса 10 с.
- **Метрики.** Успешные = только ответы 201 (`ok_messages`). Ошибки и dropped — отдельно; задержки p50/p95/p99 — только успешных. CPU — `docker stats` каждые ~1–2 с, медиана за фазу замера; CPU реплик суммируется в каждый момент. Ожидание пула — счётчики psycopg-pool из лога реплик (`STATS_LOG_SECONDS=2`). PostgreSQL — `pg_stat_database` (транзакции, WAL) и `pg_stat_statements` (операторы на сообщение). HAProxy — `show stat` (переходы в DOWN, очереди).
- **Определения.** *Потолок* конфигурации — наибольший успешный RPS по ступеням в повторе, затем медиана и разброс по 3 повторам. *SLO-ёмкость* — наибольшая предложенная нагрузка, обслуженная на ≥ 97 % с p95 ≤ 500 мс. *Узкое место* — компонент, чей CPU ≥ 90 % своего лимита на ступени потолка. Для API отдельно отмечается «одно ядро на процесс», когда CPU ≈ 100 % × число реплик при большей квоте.
### Результаты
**A. Фиксированный суммарный CPU API = 2,0 (768 МиБ), PostgreSQL 2,0 CPU** ([график](evidence/track1/A-fixed-total/baseline/throughput.svg)):
| Реплики × CPU | Потолок, успешных/с (мин–макс) | SLO-ёмкость | CPU API / квота | CPU БД / квота | CPU LB | Узкое место |
| --- | --- | --- | --- | --- | --- | --- |
| 1 × 2,0 | 214 (209–220) | 200 | 101 % / 200 % | 47 % / 200 % | 9 % | одно ядро на процесс (GIL) |
| 2 × 1,0 | **289** (281–291) | 250 | 198 % / 200 % | 66 % / 200 % | 14 % | CPU API |
| 4 × 0,5 | 239 (237–241) | 150 | 195 % / 200 % | 60 % / 200 % | 13 % | CPU API + флаппинг проверок |
Гипотеза 1 подтверждена. Одна реплика с квотой 2,0 использует ~1 ядро, вторая половина квоты простаивает. Две реплики по 1,0 используют весь бюджет и дают +35 %. Четыре по 0,5 при том же бюджете хуже двух (−17 %): процесс с квотой 0,5 обрабатывает каждый запрос вдвое дольше, растёт задержка (p95 уже при 150/с — 460 мс). Кроме того, при перегрузке реплики не успевают ответить на `/health/ready` за 1 с: HAProxy выводит их из ротации — 78 переходов в DOWN за серию, до 11 % ответов 503 от LB. Дробление ниже ядра на процесс ухудшает и скорость, и стабильность.
**B. Рост ресурсов API (1, 2, 3 реплики × 1,0 CPU) при узкой БД (PostgreSQL 0,5 CPU)** ([график](evidence/track1/B-replicas-vs-db/baseline/throughput.svg)):
| Реплики | Потолок, успешных/с (мин–макс) | CPU API / квота | CPU БД / квота | Узкое место | Ошибки при перегрузке |
| --- | --- | --- | --- | --- | --- |
| 1 | 178 (177–180) | 94 % / 100 % | 39 % / 50 % | CPU API | до 26 % |
| 2 | **210** (208–222) | 145 % / 200 % | 49 % / 50 % | CPU PostgreSQL | до 20 % |
| 3 | 197 (197–205) | 167 % / 300 % | 49 % / 50 % | CPU PostgreSQL | до 42 % |
Точка насыщения БД — около 200–210 успешных/с. Вторая реплика ещё помогает (+18 %), третья — нет: API простаивает (167 % из 300 %), CPU БД на квоте. При перегрузке выше этой точки запросы ждут соединения в пуле (до ~1 с), часть упирается в дедлайн 3 с (503). Readiness реплик тоже начинает отвечать дольше 1 с, и HAProxy выводит их из ротации (103 перехода в DOWN у 3 реплик). Больше реплик — больше конкурирующих соединений к той же БД и больше ошибок, а не больше пропускной способности.
**C. Рост ресурсов БД при 3 репликах × 1,0** ([график](evidence/track1/C-db-growth/baseline/throughput.svg)):
| CPU PostgreSQL | Потолок, успешных/с (мин–макс) | CPU API / квота | CPU БД / квота | Узкое место |
| --- | --- | --- | --- | --- |
| 0,5 (из серии B) | 197 (197–205) | 167 % / 300 % | 49 % / 50 % | PostgreSQL |
| 1,0 | **334** (316–337) | 281 % / 300 % | 94 % / 100 % | API и PostgreSQL |
| 2,0 | 347 (345–347) | 293 % / 300 % | 92 % / 200 % | CPU API |
Гипотеза 2 подтверждена причинно. Удвоение CPU одной только БД при тех же трёх репликах поднимает потолок на 70 % — значит, в B ограничивала именно БД. Следующее удвоение даёт только +4 %: узкое место вернулось в API (293 % из 300 %), БД загружена меньше чем на половину квоты. Узкое место переходит между компонентами (API → БД → API), и масштабировать имеет смысл только тот компонент, который сейчас насыщен.
**LB и пул (гипотеза 3).** HAProxy на всех ступенях занимает 4–24 % одного ядра. Очередь в LB появляется только на ступенях перегрузки, когда у реплики уже 200 одновременных запросов (`maxconn 200` на сервер): `qmax` 200–576 при 250–400/с, ни разу ниже потолка конфигурации. Это защита реплики, а не причина насыщения; `qmax` — максимум с момента старта стенда. Пул (10 соединений на реплику) не ограничивает сам по себе: ожидание соединения вырастает (сотни мс) только после насыщения CPU API или БД, то есть это симптом, а не причина. Под перегрузкой первым «ломается» не LB, а readiness: проверка `/health/ready` идёт через тот же пул и event loop, и под перегрузкой реплика честно сообщает «не готова». В итоге LB сбрасывает избыточную нагрузку ответами 503 — это видно в B и A (4 × 0,5).
### Улучшение: идемпотентная запись одним вызовом
**Диагностика.** Профиль одной реплики (2 × 1,0 CPU, 240 req/s, py-spy 20 с, [profile-baseline-3f45eb1.txt](evidence/track1/profile-baseline-3f45eb1.txt)) показал: **52,7 %** сэмплов — в драйвере psycopg, 23 % — в event loop, 9,6 % — в нашем коде, по 2–4 % — в uvicorn, redis-py, логировании, Starlette. Время делят функции, через которые проходит **каждый** оператор: ожидание ответа (`wait_async`), отправка команды, подготовка prepared statement, разбор результата. На одно сообщение старый путь делает ~15 операторов в двух транзакциях (`pg_stat_statements`): проверка членства, BEGIN, захват ключа, часы чата, вставка, COMMIT, проверки соединений пулом.
**Изменение** (коммит `4830be4`, [0003_post_message_function.sql](migrations/0003_post_message_function.sql), [repo.py](src/messenger/repo.py)). Вся логика — та же проверка членства, тот же `INSERT … ON CONFLICT`, те же часы чата и вставка — перенесена в PL/pgSQL-функцию `post_message`. Её вызывает один оператор в autocommit, вызов и есть транзакция. Семантика не меняется: оба пути проходят одни и те же тесты идемпотентности, контракта и БД (105 тестов в режиме `multi`, весь набор — в режиме `function`).
**A/B на одном образе одной переменной** (серия D, `MESSAGE_WRITE_PATH=multi|function`, [график](evidence/track1/D-improvement/4830be4/throughput.svg); пунктир — старый путь, сплошная — новый):
| Конфигурация | `multi`, потолок (мин–макс) | `function`, потолок (мин–макс) | Прирост | p95 на 450/с |
| --- | --- | --- | --- | --- |
| 2 × 1,0 CPU, БД 2,0 (упор в API) | 283 (281–286) | **489** (488–491) — не насыщен на сетке до 500/с | ≥ +73 % | `multi` — перегружен; `function` — 399 мс |
| 3 × 1,0 CPU, БД 0,5 (упор в БД) | 202 (200–203) | **452** (444–462) | +124 % | `multi` — перегружен; `function` — 331 мс |
Цена одного сообщения на ненасыщенных ступенях 200–250/с (медиана 3 повторов):
| На одно сообщение | `multi` | `function` | Изменение |
| --- | --- | --- | --- |
| CPU API | 7,2–8,6 мс | 4,3–4,9 мс | −38…44 % |
| CPU PostgreSQL | 2,3–2,6 мс | 1,0–1,3 мс | −50…60 % |
| SQL-операторов (pg_stat_statements) | 15 | 3 | −80 % |
| транзакций в pg_stat_database (с проверками пула) | ~6 | ~3 | −50 % |
| WAL | ~1,9 КБ | ~1,9 КБ | без изменений: пишутся те же строки, durability та же |
Профиль после изменения при тех же 240 req/s ([profile-function-4830be4.txt](evidence/track1/profile-function-4830be4.txt)): доля psycopg **19,0 %** вместо 52,7 %. Активных сэмплов реплики за те же 20 с — 2026 вместо 3889 (py-spy не считает простой), то есть −48 % CPU-времени на тот же поток запросов.
Механизм подтверждён измерением, а не только итоговым RPS: работа с данными (WAL) не изменилась, исчезли накладные расходы на операторы. Поэтому улучшение не зависит от конкретного ноутбука. Число операторов и транзакций на сообщение — свойство кода, оно воспроизводится на любом хосте через `pg_stat_statements`. Абсолютные RPS на другом железе будут другими.
**Потолок нового пути** (серия E, те же конфигурации, ступени 450–850/с, 3 повтора, [график](evidence/track1/E-improved-ceiling/4830be4/throughput.svg)):
| Конфигурация | `function`, потолок (мин–макс) | CPU API / квота | CPU БД / квота | Узкое место | Итог против `multi` |
| --- | --- | --- | --- | --- | --- |
| 2 × 1,0 CPU, БД 2,0 | **552** (530–552) | 195 % / 200 % | 66 % / 200 % | CPU API | 283 → 552: **+95 %** |
| 3 × 1,0 CPU, БД 0,5 | 423 (395–440) | 177 % / 300 % | 50 % / 50 % | CPU PostgreSQL | 202 → 423–452: **+110…124 %** |
В серии E DB-bound конфигурация начинает сразу со ступени 450 — это уже выше её потолка, поэтому её максимум здесь ниже, чем в D (452 при 450/с). Обе серии сходятся на ~420–460/с.
После улучшения узкие места остались на своих местах, но каждый компонент делает вдвое больше полезной работы на единицу CPU. В 2 × 1,0 потолок снова задаёт CPU API (два процесса по ядру), и следующий рост — снова реплики и ядра под API. В 3 × 1,0 при БД 0,5 потолок задаёт CPU PostgreSQL: прирост реплик API за пределами точки насыщения БД по-прежнему не помогает. Улучшение сдвинуло эту точку более чем вдвое.
### Ограничения трека
- Один хост и SMT: 8 vCPU — это 4 физических ядра. k6, LB, БД и API разнесены по `cpuset`, но делят физические ядра попарно и общий кэш, а Windows-хост создаёт фоновый шум. Абсолютные числа годятся только для сравнения конфигураций **на этом стенде** (одинаковые ресурсы, одинаковый профиль), универсального RPS нет.
- Профиль нагрузки — только запись сообщений (новые ключи, 64 чата). Чтение истории, повторы ключей и вход (Argon2id) не нагружались. Профиль с горячим чатом (все пишут в один чат) упрётся в блокировку строки чата, а не в CPU — это отдельный эксперимент.
- Сетка ступеней шагом 50/с. Потолок известен с точностью до ступени; для `function` в 2 × 1,0 потолок в серии D не достигнут, его ищет серия E.
- CPU из `docker stats` — медиана выборок ~1–2 с. Это не точный учёт cgroup за окно. Окно замера определяется по времени выхода k6 (±1–2 с).
- Замер 20 с после прогрева 10 с на пустой БД: эффекты роста таблиц (autovacuum, раздувание индексов) не исследовались.
- py-spy с `--nonblocking` слегка замедляет профилируемую реплику, поэтому профиль показывает доли, а не абсолютный RPS.
## Ограничения и следующие шаги
TODO: какие отказы выдерживаются и какие нет; что не реализовано; где учебное упрощение; какое улучшение подтверждено, а какое пока гипотеза.
**Какие отказы выдерживаются (доказано на стенде):**
- остановка, SIGKILL или заморозка **одной API-реплики** из двух: запись продолжается, подтверждённые сообщения на месте, дублей нет;
- потеря одной репликой связи с PostgreSQL/Valkey: реплика выходит из ротации, после восстановления сама возвращается;
- рестарт всего стенда с сохранением volumes: данные и прежний токен на месте, LB находит новые адреса реплик;
- поочерёдная замена реплик (rolling update) под нагрузкой — без ошибок;
- отказ PostgreSQL или Valkey целиком: ограниченные по времени 503, без анонимного доступа; после восстановления всё работает (перенесено из lab2).
**Какие не выдерживаются (честно):**
- **Один Docker-host** — это SPOF всего стенда: потеря хоста = потеря всего.
- **LB один.** Отказ HAProxy = недоступность API, хотя реплики живы. Нужны два LB с VIP (keepalived) или DNS/anycast. В Compose на одном хосте это не даёт реальной устойчивости.
- **PostgreSQL и Valkey не реплицированы.** Их отказ — 503 до восстановления, их диск — единственная копия данных (резервное копирование и PITR — тема трека 2 lab2, здесь не повторялось).
- **Обнаружение замороженной реплики** (`docker pause`) — до ~3,5 с. Запросы, попавшие на неё раньше, ждут клиентский таймаут: в доказательной серии — 4 неудачные попытки, последняя через 3,1–3,2 с (в предварительном прогоне — 8 попыток, последняя через 6,2 с: клиент дважды подряд попал на замороженную реплику до её выключения). HAProxy не повторяет такие запросы сам — они могли дойти до API, а неидемпотентные POST повторять нельзя.
- **Под глобальной перегрузкой** readiness всех реплик может стать 503 (проверка идёт через тот же пул и event loop), и LB отвечает 503 на всё. Это сброс нагрузки, но грубый. Следующий шаг — ограничение одновременных запросов на реплику ниже `maxconn` LB и отдельный лёгкий канал для readiness. Но рецензент требует исключать реплику с заблокированным пулом, поэтому зависимость readiness от пула сохранена сознательно.
- `X-Instance-Id` = hostname контейнера: в lab7 его нужно заменить непрозрачным alias.
**Учебные упрощения:** HTTP без TLS (lab7), один `uvicorn`-воркер на контейнер (масштабирование — репликами), секреты — файлы Docker secrets на одном хосте, нет rate limiting. Ключи идемпотентности хранятся в PostgreSQL рядом с данными (так проще атомарность), очистка — пачками по расписанию.
**Подтверждено измерениями:** улучшение записи сообщения (серии D и E, по 3 повтора): −80 % SQL-операторов, −40 % CPU API и −55 % CPU БД на сообщение, потолок +95 % в конфигурации с упором в API и +110…124 % с упором в БД; узкое место после улучшения — снова CPU API и CPU БД соответственно. **Пока гипотеза:** что ограничение одновременных запросов на реплику (ниже `maxconn` LB) уберёт флаппинг readiness под перегрузкой; что несколько uvicorn-воркеров в одном контейнере дадут то же, что реплики того же суммарного CPU.
+141
View File
@@ -0,0 +1,141 @@
#!/usr/bin/env python3
"""Level 4: stop EACH API replica in turn under load (restoring the previous one first) and
measure RPS, errors and recovery. Modes: graceful `stop` (SIGTERM + drain), `kill` (SIGKILL)
and `pause` (frozen process: the kernel still accepts TCP, nothing answers - the worst case
for detection, only the active health check can find it).
python bench/failover_series.py --output evidence/failover # ~5 min for 2 replicas x 3 modes
Timeline of one round (all times from the moment the stop command is issued):
warm-up 6 s -> stop replica -> 12 s with it down -> start it again -> wait until HAProxy
has every replica UP (rejoin time) -> 6 s steady state.
Measured per round: requests/ok/failed and ok-RPS before / during / after, the last failed
request (+s), the first moment after which the stopped replica never answered again
(= removal from rotation), time until the restarted replica is UP in HAProxy and serves
traffic again, plus the history check (no confirmed message lost, no duplicate).
Output: <output>/series.json (summary), <output>/timeline-<mode>-<replica>.csv (per second),
<output>/series.md (table for REPORT.md). No tokens are written.
"""
from __future__ import annotations
import argparse
import csv
import datetime as dt
import json
import platform
import subprocess
import sys
import time
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT / "scripts"))
sys.path.insert(0, str(ROOT / "bench"))
import lbctl # noqa: E402
from loadgen import Load # noqa: E402
def docker(*args: str) -> str:
return subprocess.run(["docker", *args], capture_output=True, text=True, check=True, timeout=120).stdout
def one_round(replica: dict, mode: str, replicas: int, workers: int, out: Path) -> dict:
load = Load(workers).start()
time.sleep(6)
t_stop = load.now()
if mode == "pause":
docker("pause", replica["id"])
else:
docker("stop", "--time", "15" if mode == "stop" else "0", replica["id"])
t_stopped = load.now()
time.sleep(12)
t_start = load.now()
docker("unpause" if mode == "pause" else "start", replica["id"])
lbctl.wait_up(replicas, timeout=90)
t_up = load.now()
# The restarted container keeps its hostname: wait until it serves requests again.
deadline = time.monotonic() + 30
while time.monotonic() < deadline:
if any(e[3] == replica["hostname"] and e[0] > t_up for e in load.events):
break
time.sleep(0.1)
t_serves = load.now()
time.sleep(6)
load.stop()
t_end = load.now()
failed_after_stop = [e for e in load.events if e[0] >= t_stop and e[2] not in (200, 201)]
served_by_target = [e[0] for e in load.events if e[3] == replica["hostname"] and t_stop <= e[0] < t_start]
out_of_rotation = (max(served_by_target) - t_stop) if served_by_target else 0.0
rows = load.per_second()
with (out / f"timeline-{mode}-{replica['name']}.csv").open("w", newline="", encoding="utf-8") as f:
writer = csv.DictWriter(f, fieldnames=["second", "ok", "failed", "instances"])
writer.writeheader()
for row in rows:
row = dict(row)
row["instances"] = row["instances"].replace(replica["hostname"], "TARGET")
writer.writerow(row)
state = json.loads(docker("inspect", replica["id"]))[0]["State"]
return {
"mode": mode, "replica": replica["name"], "instance_id": replica["hostname"],
"stop_command_seconds": round(t_stopped - t_stop, 2),
"before": load.window(0.5, t_stop), "during": load.window(t_stop, t_start), "after": load.window(t_serves, t_end),
"failed_after_stop": len(failed_after_stop),
"failed_statuses": sorted({e[2] for e in failed_after_stop}),
"last_failure_after_stop_s": round(max((e[0] for e in failed_after_stop), default=t_stop) - t_stop, 3),
"target_last_served_after_stop_s": round(out_of_rotation, 3),
"rejoin_up_in_lb_s": round(t_up - t_start, 2), "rejoin_serving_s": round(t_serves - t_start, 2),
"client_retries_same_key": load.retries,
"history": load.verify_history(),
"restarted_state": f"{state['Status']} health={(state.get('Health') or {}).get('Status')}",
}
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
parser.add_argument("--output", default="evidence/failover")
parser.add_argument("--modes", default="stop,kill,pause")
parser.add_argument("--workers", type=int, default=4)
args = parser.parse_args()
out = ROOT / args.output
out.mkdir(parents=True, exist_ok=True)
replicas = sorted((r for r in lbctl.replicas() if r["state"] == "running"), key=lambda r: r["name"])
assert len(replicas) >= 2, "need at least two running replicas"
lbctl.wait_up(len(replicas))
results = []
for mode in args.modes.split(","):
for replica in replicas: # each replica in turn; the previous one is already restored
print(f"== {mode} {replica['name']} ...", flush=True)
result = one_round(replica, mode, len(replicas), args.workers, out)
results.append(result)
print(json.dumps({k: result[k] for k in ("failed_after_stop", "last_failure_after_stop_s",
"target_last_served_after_stop_s", "rejoin_serving_s")}), flush=True)
meta = {"date": dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"),
"commit": subprocess.run(["git", "rev-parse", "--short", "HEAD"], cwd=ROOT, capture_output=True,
text=True).stdout.strip(),
"host": f"{platform.system()} {platform.release()}, docker {docker('version', '--format', '{{.Server.Version}}').strip()}",
"workload": f"{args.workers} closed-loop workers from the host via 127.0.0.1:8080: POST message "
"(Idempotency-Key, uncertain -> same key) + GET last page; client timeout 3 s",
"lb": "HAProxy: check GET /health/ready inter 1s fastinter 500ms fall 2 rise 2, timeout check 1s; "
"retry-on conn-failure + redispatch",
"api": "DRAIN_SECONDS=3, SHUTDOWN_TIMEOUT_SECONDS=8, stop_grace_period 15s"}
(out / "series.json").write_text(json.dumps({"meta": meta, "rounds": results}, ensure_ascii=False, indent=2) + "\n",
encoding="utf-8")
lines = ["| Режим | Реплика | Запросов во время отказа | Ошибок после stop | ok-RPS до / во время / после | "
"Последняя ошибка, с | Цель ушла из ротации, с | Возврат в LB / обслуживает, с | Потери / дубли |",
"| --- | --- | --- | --- | --- | --- | --- | --- | --- |"]
for r in results:
h = r["history"]
lines.append(f"| {r['mode']} | {r['replica']} | {r['during']['requests']} | {r['failed_after_stop']} "
f"{r['failed_statuses'] or ''} | {r['before']['ok_rps']} / {r['during']['ok_rps']} / {r['after']['ok_rps']} | "
f"{r['last_failure_after_stop_s']} | {r['target_last_served_after_stop_s']} | "
f"{r['rejoin_up_in_lb_s']} / {r['rejoin_serving_s']} | "
f"{h['lost_confirmed']} / {h['duplicate_ids'] + h['duplicate_texts']} |")
(out / "series.md").write_text("\n".join([f"Дата {meta['date']}, commit {meta['commit']}. {meta['workload']}.", "",
*lines, ""]), encoding="utf-8")
print("\n".join(lines))
return 0
if __name__ == "__main__":
raise SystemExit(main())
+212
View File
@@ -0,0 +1,212 @@
#!/usr/bin/env python3
"""Background HTTP traffic through the LB for failover / rolling-update experiments (level 4).
Closed loop, N worker threads, each: POST a message (Idempotency-Key) + GET the last page.
An uncertain POST (timeout, connection error, 5xx) is retried with the SAME key and body,
exactly as a correct client must do; the history is checked afterwards for losses and
duplicates. Every request is recorded as (t, kind, status, instance, latency).
This is a correctness/availability probe, not a capacity benchmark (see bench/track1 for that).
python bench/loadgen.py --duration 30 --workers 4 # prints a summary JSON
"""
from __future__ import annotations
import argparse
import base64
import json
import math
import os
import threading
import time
import uuid
from collections import Counter, defaultdict
import urllib.error
import urllib.request
BASE_URL = os.environ.get("BASE_URL", "http://127.0.0.1:8080").rstrip("/")
class Http:
"""Minimal stdlib client with keep-alive per thread (http.client under urllib is not)."""
def __init__(self, base_url: str, timeout: float):
import http.client
from urllib.parse import urlsplit
parts = urlsplit(base_url)
self.host, self.port = parts.hostname, parts.port or 80
self.timeout = timeout
self.factory = lambda: http.client.HTTPConnection(self.host, self.port, timeout=self.timeout)
self.conn = self.factory()
def request(self, method: str, path: str, body: bytes | None = None, headers: dict | None = None):
try:
self.conn.request(method, path, body=body, headers=headers or {})
response = self.conn.getresponse()
data = response.read()
if response.getheader("connection", "").lower() == "close":
self.conn.close()
self.conn = self.factory()
return response.status, dict(response.getheaders()), data
except Exception:
self.conn.close()
self.conn = self.factory()
raise
def call(method: str, path: str, body=None, headers=None, timeout=10.0):
data = json.dumps(body).encode() if body is not None else None
h = {"Content-Type": "application/json", **(headers or {})}
req = urllib.request.Request(BASE_URL + path, data=data, headers=h, method=method)
try:
with urllib.request.urlopen(req, timeout=timeout) as r:
return r.status, json.loads(r.read() or b"null")
except urllib.error.HTTPError as e:
return e.code, json.loads(e.read() or b"null")
def make_user(label: str) -> dict:
username = f"{label}_{uuid.uuid4().hex[:12]}"
password = "Load-password-" + uuid.uuid4().hex
status, user = call("POST", "/api/v1/users", {"username": username, "display_name": label, "password": password})
assert status == 201, (status, user)
cred = base64.b64encode(f"{username}:{password}".encode()).decode()
status, session = call("POST", "/api/v1/auth/sessions", headers={"Authorization": "Basic " + cred})
assert status == 201, (status, session)
return {"id": user["id"], "token": session["token"]}
class Load:
def __init__(self, workers: int = 4, timeout: float = 3.0, pause: float = 0.0):
self.workers = workers
self.timeout = timeout
self.pause = pause
self.alice, self.bob = make_user("load"), make_user("load")
status, chat = call("POST", "/api/v1/chats", {"title": "load", "member_ids": [self.alice["id"], self.bob["id"]]},
headers={"Authorization": "Bearer " + self.alice["token"]})
assert status == 201, chat
self.chat = chat
self.stop_event = threading.Event()
self.lock = threading.Lock()
self.events: list[tuple[float, str, int, str, float]] = []
self.confirmed: dict[str, dict] = {}
self.retries = 0
self.t0 = time.monotonic()
self.threads = [threading.Thread(target=self._worker, args=(i,), daemon=True) for i in range(workers)]
def now(self) -> float:
return time.monotonic() - self.t0
def _record(self, kind: str, status: int, instance: str, latency: float) -> None:
with self.lock:
self.events.append((round(self.now(), 4), kind, status, instance, round(latency * 1000, 2)))
def _worker(self, index: int) -> None:
http = Http(BASE_URL, self.timeout)
auth = {"Authorization": "Bearer " + self.alice["token"]}
path = f"/api/v1/chats/{self.chat['id']}/messages"
n, key = 0, uuid.uuid4().hex
text = f"w{index}-0-{key[:6]}"
while not self.stop_event.is_set():
started = time.monotonic()
try:
status, headers, data = http.request("POST", path, json.dumps({"text": text}).encode(),
{**auth, "Idempotency-Key": key, "Content-Type": "application/json"})
instance = headers.get("x-instance-id") or headers.get("X-Instance-Id") or "-"
except Exception:
status, instance, data = 0, "-", b""
self._record("POST", status, instance, time.monotonic() - started)
if status == 201:
msg = json.loads(data)
with self.lock:
self.confirmed[msg["id"]] = msg
n += 1
key = uuid.uuid4().hex
text = f"w{index}-{n}-{key[:6]}"
elif status in (0, 502, 503, 504, 500):
with self.lock:
self.retries += 1 # same key + same text on the next iteration
started = time.monotonic()
try:
status, headers, _ = http.request("GET", path + "?limit=5", None, auth)
instance = headers.get("x-instance-id") or headers.get("X-Instance-Id") or "-"
except Exception:
status, instance = 0, "-"
self._record("GET", status, instance, time.monotonic() - started)
if self.pause:
time.sleep(self.pause)
def start(self) -> "Load":
self.t0 = time.monotonic()
for t in self.threads:
t.start()
return self
def stop(self) -> None:
self.stop_event.set()
for t in self.threads:
t.join(timeout=15)
def verify_history(self) -> dict:
"""Walk the whole history with the second user: no confirmed message lost, no duplicate."""
items, cursor = [], None
while True:
q = "?limit=100" + (f"&cursor={cursor}" if cursor else "")
status, page = call("GET", f"/api/v1/chats/{self.chat['id']}/messages{q}",
headers={"Authorization": "Bearer " + self.bob["token"]}, timeout=30)
assert status == 200, page
items += page["items"]
cursor = page["next_cursor"]
if not cursor:
break
ids = [m["id"] for m in items]
texts = [m["text"] for m in items]
lost = [i for i in self.confirmed if i not in set(ids)]
changed = [m for m in items if m["id"] in self.confirmed and self.confirmed[m["id"]] != m]
return {"history_messages": len(items), "confirmed_messages": len(self.confirmed),
"lost_confirmed": len(lost), "altered_confirmed": len(changed),
"duplicate_ids": len(ids) - len(set(ids)), "duplicate_texts": len(texts) - len(set(texts)),
"unconfirmed_but_stored": len(set(ids) - set(self.confirmed))}
# ------------------------------------------------------------------ summaries
def window(self, start: float, end: float) -> dict:
ev = [e for e in self.events if start <= e[0] < end]
ok = [e for e in ev if e[2] in (200, 201)]
lat = sorted(e[4] for e in ok)
def pct(p: float):
return lat[max(0, math.ceil(p * len(lat)) - 1)] if lat else None
span = max(end - start, 1e-9)
return {"seconds": round(span, 2), "requests": len(ev), "ok": len(ok), "failed": len(ev) - len(ok),
"ok_rps": round(len(ok) / span, 1), "error_ratio": round((len(ev) - len(ok)) / len(ev), 5) if ev else None,
"statuses": dict(Counter(e[2] for e in ev)), "instances": dict(Counter(e[3] for e in ok)),
"p50_ms": pct(.5), "p95_ms": pct(.95), "p99_ms": pct(.99)}
def per_second(self) -> list[dict]:
buckets: dict[int, list] = defaultdict(list)
for e in self.events:
buckets[int(e[0])].append(e)
rows = []
for second in sorted(buckets):
ev = buckets[second]
rows.append({"second": second, "ok": sum(1 for e in ev if e[2] in (200, 201)),
"failed": sum(1 for e in ev if e[2] not in (200, 201)),
"instances": "|".join(f"{k}:{v}" for k, v in sorted(Counter(e[3] for e in ev).items()))})
return rows
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
parser.add_argument("--duration", type=float, default=20)
parser.add_argument("--workers", type=int, default=4)
args = parser.parse_args()
load = Load(args.workers).start()
time.sleep(args.duration)
load.stop()
print(json.dumps({"total": load.window(0, load.now()), "history": load.verify_history(),
"client_retries_same_key": load.retries}, ensure_ascii=False, indent=2))
return 0
if __name__ == "__main__":
raise SystemExit(main())
+75
View File
@@ -0,0 +1,75 @@
#!/usr/bin/env python3
"""Level 4: scripts/rolling-update.sh while bench/loadgen.py traffic runs through the LB.
python bench/rolling_under_load.py --output evidence/rolling-update
Writes <output>/summary.json (errors, RPS, history check, revisions before/after),
<output>/timeline.csv (per second ok/failed/instances) and <output>/script-output.txt.
"""
from __future__ import annotations
import argparse
import csv
import json
import os
import subprocess
import sys
import time
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT / "scripts"))
sys.path.insert(0, str(ROOT / "bench"))
import lbctl # noqa: E402
from loadgen import Load # noqa: E402
def revisions() -> dict[str, str]:
result = {}
for r in lbctl.replicas():
if r["state"] == "running":
label = subprocess.run(["docker", "inspect", "-f", '{{index .Config.Labels "org.opencontainers.image.revision"}}',
r["id"]], capture_output=True, text=True).stdout.strip()
result[r["name"]] = f"{r['hostname']} revision={label}"
return result
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
parser.add_argument("--output", default="evidence/rolling-update")
parser.add_argument("--workers", type=int, default=4)
args = parser.parse_args()
out = ROOT / args.output
out.mkdir(parents=True, exist_ok=True)
before = revisions()
load = Load(args.workers).start()
time.sleep(5)
t_begin = load.now()
bash = (os.path.join(os.environ.get("ProgramFiles", r"C:\Program Files"), "Git", "bin", "bash.exe")
if os.name == "nt" else "sh")
env = {**os.environ, "PYTHON": sys.executable}
script = subprocess.run([bash, str(ROOT / "scripts" / "rolling-update.sh")], cwd=ROOT, env=env,
capture_output=True, text=True, encoding="utf-8", timeout=900)
t_end = load.now()
time.sleep(5)
load.stop()
(out / "script-output.txt").write_text(f"$ sh scripts/rolling-update.sh (exit {script.returncode})\n"
+ script.stdout + script.stderr, encoding="utf-8")
with (out / "timeline.csv").open("w", newline="", encoding="utf-8") as f:
writer = csv.DictWriter(f, fieldnames=["second", "ok", "failed", "instances"])
writer.writeheader()
writer.writerows(load.per_second())
summary = {"script_exit_code": script.returncode, "update_seconds": round(t_end - t_begin, 1),
"before": before, "after": revisions(),
"during_update": load.window(t_begin, t_end), "whole_run": load.window(0, load.now()),
"client_retries_same_key": load.retries, "history": load.verify_history(),
"workload": f"{args.workers} closed-loop workers: POST (Idempotency-Key) + GET page via 127.0.0.1:8080"}
(out / "summary.json").write_text(json.dumps(summary, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
print(json.dumps({k: summary[k] for k in ("script_exit_code", "update_seconds", "before", "after", "history")},
ensure_ascii=False, indent=2))
print(json.dumps(summary["during_update"], ensure_ascii=False))
return 0 if script.returncode == 0 else 1
if __name__ == "__main__":
raise SystemExit(main())
+220
View File
@@ -0,0 +1,220 @@
#!/usr/bin/env python3
"""Track 1 analysis: evidence/track1/<series>/<label>/results.csv -> summary.md + SVG charts.
python bench/track1/analyze.py # all series found under evidence/track1
Per configuration and offered rate: median over repeats (and min..max) of SUCCESSFUL req/s
(only 201 responses count; errors and dropped iterations are failed demand), p95 of successful
requests, CPU of every component as a share of ITS OWN limit. Capacity of a configuration =
per repeat the highest successful req/s over the sweep, then median and range over repeats.
SLO capacity = the highest offered rate that was served: ok >= 97 % of offered, p95 <= 500 ms.
Bottleneck = the component whose CPU is >= 90 % of its limit at the capacity step.
Standard library only; charts are plain SVG (one y-axis, <= 4 series, legend + labels, hover titles).
"""
from __future__ import annotations
import csv
import re
import statistics
from collections import defaultdict
from pathlib import Path
ROOT = Path(__file__).resolve().parents[2]
EVIDENCE = ROOT / "evidence" / "track1"
SERIES_COLORS = [("#2a78d6", "#3987e5"), ("#eb6834", "#d95926"), ("#1baf7a", "#199e70"), ("#eda100", "#c98500")]
def num(value):
if value in (None, "", "None"):
return None
return float(value)
CONFIG_RE = re.compile(r"n(?P<n>\d+)-api(?P<api>[\d.]+)-pg(?P<pg>[\d.]+)-pool(?P<pool>\d+)(?:-message_write_path_(?P<path>\w+))?")
def label_of(config: str) -> str:
m = CONFIG_RE.fullmatch(config)
if not m:
return config
text = f"{m['n']} × {m['api']} CPU API, БД {m['pg']} CPU"
if m["path"]:
text += f", запись {m['path']}"
return text
def load(path: Path) -> list[dict]:
with path.open(encoding="utf-8") as f:
return list(csv.DictReader(f))
def aggregate(rows: list[dict]) -> dict:
by_cfg: dict[str, dict] = defaultdict(lambda: defaultdict(list))
for r in rows:
by_cfg[r["config"]][int(r["rate"])].append(r)
result = {}
for cfg, by_rate in by_cfg.items():
points = []
for rate in sorted(by_rate):
rs = by_rate[rate]
ok = [num(r["ok_rps"]) for r in rs]
p95 = [num(r["p95_ms"]) for r in rs if num(r["p95_ms"]) is not None]
points.append({"rate": rate, "ok": statistics.median(ok), "lo": min(ok), "hi": max(ok), "n": len(rs),
"p95": statistics.median(p95) if p95 else None,
"err": statistics.median(num(r["error_ratio"]) for r in rs),
"dropped": statistics.median(num(r["dropped"]) for r in rs),
"api_cpu": statistics.median(num(r["api_cpu_pct_total"]) or 0 for r in rs),
"pg_cpu": statistics.median(num(r["pg_cpu_pct"]) or 0 for r in rs),
"lb_cpu": statistics.median(num(r["lb_cpu_pct"]) or 0 for r in rs),
"k6_cpu": statistics.median(num(r["k6_cpu_pct"]) or 0 for r in rs),
"pool_wait": statistics.median(num(r["pool_wait_ms_avg"]) or 0 for r in rs),
"stmts": statistics.median(num(r["statements_per_ok"]) or 0 for r in rs),
"commits": statistics.median(num(r["pg_commits_per_ok"]) or 0 for r in rs),
"chkdown": sum(int(num(r["lb_chkdown"]) or 0) for r in rs),
"api_limit": num(rs[0].get("api_cpus_limit_total")), "pg_limit": num(rs[0].get("pg_cpus_limit")),
"replicas": num(rs[0].get("replicas_running"))})
repeats = sorted({int(r["repeat"]) for rr in by_rate.values() for r in rr})
caps, slo = [], []
for rep in repeats:
rr = [r for rate_rows in by_rate.values() for r in rate_rows if int(r["repeat"]) == rep]
caps.append(max(num(r["ok_rps"]) for r in rr))
served = [int(r["rate"]) for r in rr if num(r["ok_rps"]) >= 0.97 * int(r["rate"])
and (num(r["p95_ms"]) or 1e9) <= 500]
slo.append(max(served) if served else 0)
best = max(points, key=lambda p: p["ok"])
bottleneck = []
if best["api_limit"] and best["api_cpu"] >= 90 * best["api_limit"]:
bottleneck.append("API CPU")
elif best["replicas"] and best["api_cpu"] >= 90 * best["replicas"] and best["api_limit"] > best["replicas"]:
# One uvicorn process per replica runs Python bytecode on ONE core (GIL), whatever the quota.
bottleneck.append("API: 1 ядро на процесс (GIL), квота не используется")
if best["pg_limit"] and best["pg_cpu"] >= 90 * best["pg_limit"]:
bottleneck.append("PostgreSQL CPU")
if best["lb_cpu"] >= 90:
bottleneck.append("LB CPU")
result[cfg] = {"points": points, "capacity": statistics.median(caps), "cap_lo": min(caps), "cap_hi": max(caps),
"slo": statistics.median(slo), "repeats": len(repeats), "best": best,
"bottleneck": ", ".join(bottleneck) or "нет (ни один CPU не у предела)"}
return result
def svg_chart(title: str, series: dict[str, list[dict]], path: Path, y_key: str = "ok",
y_label: str = "успешных запросов/с (201)", diagonal: bool = True) -> None:
width, height, left, right, top, bottom = 760, 420, 64, 210, 40, 52
xs = [p["rate"] for pts in series.values() for p in pts]
ys = [p.get("hi", p[y_key]) for pts in series.values() for p in pts] + ([max(xs)] if diagonal else [])
x0, x1 = 0, max(xs) * 1.05
y1 = max(ys) * 1.1
sx = lambda v: left + (v - x0) / (x1 - x0) * (width - left - right) # noqa: E731
sy = lambda v: height - bottom - v / y1 * (height - top - bottom) # noqa: E731
out = [f'<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 {width} {height}" font-family="system-ui, sans-serif" font-size="12">',
"<style>.bg{fill:#fcfcfb}.ink{fill:#0b0b0b}.ink2{fill:#52514e}.grid{stroke:#e4e3df}.ref{stroke:#b5b3ac}"
+ "".join(f".l{i}{{stroke:{c[0]};fill:none}}.d{i}{{fill:{c[0]};stroke:#fcfcfb}}" for i, c in enumerate(SERIES_COLORS))
+ "@media (prefers-color-scheme: dark){.bg{fill:#1a1a19}.ink{fill:#fff}.ink2{fill:#c3c2b7}.grid{stroke:#34332f}.ref{stroke:#6b6a64}"
+ "".join(f".l{i}{{stroke:{c[1]}}}.d{i}{{fill:{c[1]};stroke:#1a1a19}}" for i, c in enumerate(SERIES_COLORS)) + "}</style>",
f'<rect class="bg" width="{width}" height="{height}"/>',
f'<text class="ink" x="{left}" y="22" font-size="14" font-weight="600">{title}</text>']
step = nice_step(y1)
v = 0.0
while v <= y1:
out.append(f'<line class="grid" x1="{left}" x2="{width - right}" y1="{sy(v):.1f}" y2="{sy(v):.1f}" stroke-width="1"/>')
out.append(f'<text class="ink2" x="{left - 8}" y="{sy(v) + 4:.1f}" text-anchor="end">{v:g}</text>')
v += step
for rate in sorted(set(xs)):
out.append(f'<text class="ink2" x="{sx(rate):.1f}" y="{height - bottom + 18}" text-anchor="middle">{rate}</text>')
out.append(f'<text class="ink2" x="{(left + width - right) / 2}" y="{height - 12}" text-anchor="middle">предложенная нагрузка, запросов/с (open-loop)</text>')
out.append(f'<text class="ink2" transform="translate(16 {(top + height - bottom) / 2}) rotate(-90)" text-anchor="middle">{y_label}</text>')
if diagonal:
out.append(f'<line class="ref" x1="{sx(0):.1f}" y1="{sy(0):.1f}" x2="{sx(max(xs)):.1f}" y2="{sy(max(xs)):.1f}" stroke-width="1"/>')
mid = min(xs) + 0.25 * (max(xs) - min(xs))
out.append(f'<text class="ink2" x="{sx(mid) - 6:.1f}" y="{sy(mid) - 8:.1f}" text-anchor="end">обслужено = предложено</text>')
legend_y = top + 10
# Colour = resource configuration; line style = write path (dashed = old `multi`), so an
# A/B pair of one configuration shares a hue and reads as "before / after".
bases = list(dict.fromkeys(name.split(", запись")[0] for name in series))
for name, pts in series.items():
i = bases.index(name.split(", запись")[0])
dash = ' stroke-dasharray="6 4"' if name.endswith("запись multi") else ""
line_cls, dot_cls = f"l{i}", f"d{i}"
line = " ".join(f"{sx(p['rate']):.1f},{sy(p[y_key]):.1f}" for p in pts)
out.append(f'<polyline class="{line_cls}" points="{line}" stroke-width="2" stroke-linejoin="round" stroke-linecap="round"{dash}/>')
for p in pts:
if "lo" in p and y_key == "ok":
out.append(f'<line class="{line_cls}" x1="{sx(p["rate"]):.1f}" x2="{sx(p["rate"]):.1f}" y1="{sy(p["lo"]):.1f}" y2="{sy(p["hi"]):.1f}" stroke-width="2" opacity="0.5"/>')
tip = (f"{name}: предложено {p['rate']}/с, успешно {p[y_key]:g}/с" + (f" (повторы {p['lo']:g}–{p['hi']:g})" if "lo" in p else "")
+ (f", p95 {p['p95']:.0f} мс" if p.get("p95") else ""))
out.append(f'<circle class="{dot_cls}" cx="{sx(p["rate"]):.1f}" cy="{sy(p[y_key]):.1f}" r="4" stroke-width="2"><title>{tip}</title></circle>')
peak = max(pts, key=lambda q: q[y_key]) # direct label = the ceiling of this series
out.append(f'<text class="ink2" x="{sx(peak["rate"]):.1f}" y="{sy(peak[y_key]) - 10:.1f}" text-anchor="middle">max {peak[y_key]:.0f}</text>')
out.append(f'<line class="{line_cls}" x1="{width - right + 34}" x2="{width - right + 52}" y1="{legend_y}" y2="{legend_y}" stroke-width="2"{dash}/>')
for k, chunk in enumerate(wrap(name, 22)):
out.append(f'<text class="ink" x="{width - right + 58}" y="{legend_y + 4 + 14 * k}">{chunk}</text>')
legend_y += 14 * len(wrap(name, 22)) + 12
out.append("</svg>")
path.write_text("\n".join(out) + "\n", encoding="utf-8")
def wrap(text: str, width: int) -> list[str]:
words, lines, cur = text.split(), [], ""
for w in words:
if len(cur) + len(w) + 1 > width and cur:
lines.append(cur)
cur = w
else:
cur = f"{cur} {w}".strip()
return lines + [cur]
def nice_step(top: float) -> float:
for step in (10, 20, 25, 50, 100, 200, 250, 500, 1000):
if top / step <= 8:
return step
return 2000
def series_section(series_dir: Path) -> list[str]:
lines = []
for label_dir in sorted(p for p in series_dir.iterdir() if (p / "results.csv").exists()):
rows = load(label_dir / "results.csv")
if not rows:
continue # series still running / not started
agg = aggregate(rows)
name = f"{series_dir.name}/{label_dir.name}"
lines += [f"### {name}", "", "| Конфигурация | Повторов | Потолок, успешных/с (медиана, мин–макс) | SLO-ёмкость, предложено/с | "
"p95 на потолке, мс | CPU API / лимит | CPU БД / лимит | CPU LB | CPU k6 | Ожидание пула, мс | SQL-операторов на сообщение | Узкое место |",
"| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |"]
for cfg, a in agg.items():
b = a["best"]
lines.append(f"| {label_of(cfg)} | {a['repeats']} | **{a['capacity']:.0f}** ({a['cap_lo']:.0f}–{a['cap_hi']:.0f}) | {a['slo']:.0f} | "
f"{b['p95']:.0f} | {b['api_cpu']:.0f} % / {100 * (b['api_limit'] or 0):.0f} % | "
f"{b['pg_cpu']:.0f} % / {100 * (b['pg_limit'] or 0):.0f} % | {b['lb_cpu']:.0f} % | {b['k6_cpu']:.0f} % | "
f"{b['pool_wait']:.1f} | {b['stmts']:.1f} | {a['bottleneck']} |")
lines += ["", "<details><summary>Все ступени (медианы по повторам)</summary>", "",
"| Конфигурация | Предложено/с | Успешно/с (мин–макс) | p95, мс | Ошибки | Dropped | CPU API | CPU БД | Ожидание пула, мс | DOWN в LB |",
"| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |"]
for cfg, a in agg.items():
for p in a["points"]:
lines.append(f"| {label_of(cfg)} | {p['rate']} | {p['ok']:.1f} ({p['lo']:.0f}–{p['hi']:.0f}) | "
f"{(p['p95'] or 0):.0f} | {p['err']:.4f} | {p['dropped']:.0f} | {p['api_cpu']:.0f} % | {p['pg_cpu']:.0f} % | "
f"{p['pool_wait']:.1f} | {p['chkdown']} |")
lines += ["", "</details>", ""]
chart = label_dir / "throughput.svg"
svg_chart(name, {label_of(c): a["points"] for c, a in agg.items()}, chart)
lines += [f"![{name}]({chart.relative_to(EVIDENCE).as_posix()})", ""]
return lines
def main() -> int:
lines = ["# Трек 1 — сводка (сгенерировано bench/track1/analyze.py из results.csv)", "",
"Успешные = только ответы 201 в фазе measure; ошибки и dropped iterations k6 — неудовлетворённый спрос. "
"CPU в % одного ядра (100 % = 1 CPU), медиана выборок docker stats за фазу measure; у API — сумма реплик в каждый "
"момент. Лимиты — фактические из docker inspect.", ""]
for series_dir in sorted(p for p in EVIDENCE.iterdir() if p.is_dir() and p.name[:1].isupper()):
lines += [f"## {series_dir.name}", ""] + series_section(series_dir)
(EVIDENCE / "summary.md").write_text("\n".join(lines) + "\n", encoding="utf-8")
print((EVIDENCE / "summary.md").read_text(encoding="utf-8")[:4000])
return 0
if __name__ == "__main__":
raise SystemExit(main())
+31
View File
@@ -0,0 +1,31 @@
# Track 1 overlay: docker compose -p mtusi-lab3-bench -f compose.yaml -f bench/track1/compose.bench.yaml ...
# Used only by bench/track1/run.py. A separate project (own volumes, own ports) so the
# experiments never touch the main stand's data. Every role is pinned to its own CPU set,
# so the load generator, the LB and PostgreSQL do not steal CPU from the API replicas and
# the "total API CPU" of a configuration is exactly what the limits say:
# k6 0-1 | lb + valkey 2 | postgres 3-4 | api 5-7 (all replicas share these 3 CPUs; per-replica quota API_CPUS)
services:
postgres:
cpuset: "${PG_CPUSET:-3-4}"
valkey:
cpuset: "${VALKEY_CPUSET:-2}"
lb:
cpuset: "${LB_CPUSET:-2}"
migrate:
cpuset: "${API_CPUSET:-5-7}"
api:
cpuset: "${API_CPUSET:-5-7}"
environment:
STATS_LOG_SECONDS: "${STATS_LOG_SECONDS:-2}"
k6:
image: grafana/k6:2.3.0@sha256:9c2dee7f8ed74d317e4027c06a10f169b625638189de8d4555d0b3486a5aeb34
profiles: ["bench"]
cpuset: "${K6_CPUSET:-0-1}"
networks: [edge] # reaches the LB only, like an external client
environment:
BASE_URL: http://lb:8080
volumes:
- ./bench/track1/k6:/scripts:ro
deploy:
resources:
limits: {cpus: "2.0", memory: 1g}
+67
View File
@@ -0,0 +1,67 @@
// Track 1 workload: OPEN-LOOP arrival rate of "send message" requests through the LB.
// Every iteration = one POST /api/v1/chats/{id}/messages with a NEW Idempotency-Key into one
// of USERS chats (user i writes into chat i, shared with user i+1), so writers are spread over
// many chat rows and no single-row lock dominates. constant-arrival-rate starts iterations at
// RATE/s regardless of response time (no coordinated omission); iterations that cannot start
// because all VUs are busy are counted in dropped_iterations - they are failed demand.
// Phases: `warmup` (not reported) then `measure`. Only status 201 counts as success.
// Env: BASE_URL, RATE, WARMUP, DURATION, MAX_VUS, CHATS_FILE.
import http from 'k6/http';
import { check } from 'k6';
import exec from 'k6/execution';
import { Counter, Trend } from 'k6/metrics';
const BASE = __ENV.BASE_URL || 'http://lb:8080';
const RATE = parseInt(__ENV.RATE || '100', 10);
const WARMUP = __ENV.WARMUP || '10s';
const DURATION = __ENV.DURATION || '20s';
const MAX_VUS = parseInt(__ENV.MAX_VUS || '400', 10);
const ok = new Counter('ok_messages'); // 201 only
const okLatency = new Trend('ok_latency', true); // latency of 201 only
function phase(name, startTime, duration) {
return {
executor: 'constant-arrival-rate', rate: RATE, timeUnit: '1s', duration, startTime,
preAllocatedVUs: Math.min(MAX_VUS, Math.max(20, Math.ceil(RATE / 10))), maxVUs: MAX_VUS,
gracefulStop: '5s', exec: 'send', tags: { phase: name },
};
}
export const options = {
discardResponseBodies: true,
scenarios: { warmup: phase('warmup', '0s', WARMUP), measure: phase('measure', WARMUP, DURATION) },
summaryTrendStats: ['avg', 'p(50)', 'p(95)', 'p(99)', 'max'],
// Thresholds only create per-phase sub-metrics in the summary; they never fail the run.
thresholds: {
'http_reqs{phase:measure}': ['count>=0'],
'http_req_failed{phase:measure}': ['rate>=0'],
'http_req_duration{phase:measure}': ['max>=0'],
'ok_messages{phase:measure}': ['count>=0'],
'ok_latency{phase:measure}': ['max>=0'],
'dropped_iterations{scenario:measure}': ['count>=0'],
'iterations{scenario:measure}': ['count>=0'],
},
};
// Users, sessions and chats are created by bench/track1/run.py before k6 starts (so the timed
// phases start at once); the file with the synthetic tokens is mounted read-only and deleted after.
const chats = JSON.parse(open(__ENV.CHATS_FILE || '/data/chats.json'));
export function send() {
const n = exec.scenario.iterationInTest;
const c = chats[n % chats.length];
const res = http.post(`${BASE}/api/v1/chats/${c.chat}/messages`,
JSON.stringify({ text: `bench ${exec.vu.idInTest}-${n}` }),
{ headers: { 'Content-Type': 'application/json', Authorization: `Bearer ${c.token}`,
'Idempotency-Key': `${exec.scenario.name}-${exec.vu.idInTest}-${n}-${Date.now()}` },
tags: { name: 'post_message' }, timeout: '10s' });
if (check(res, { '201': (r) => r.status === 201 })) {
ok.add(1);
okLatency.add(res.timings.duration);
}
}
export function handleSummary(data) {
return { stdout: '\n@@SUMMARY@@' + JSON.stringify(data) + '\n' };
}
+125
View File
@@ -0,0 +1,125 @@
#!/usr/bin/env python3
"""Where does one API replica spend its CPU under the track 1 workload? (py-spy sidecar)
python bench/track1/profile.py --label baseline --rate 260 --config replicas=2,api_cpus=1.0
Starts a fresh bench stand, runs k6 at a fixed arrival rate and samples ONE replica with
py-spy (sidecar sharing its PID namespace, SYS_PTRACE only for the sidecar, image
bench/track1/pyspy with a hash-pinned py-spy). Output evidence/track1/profile-<label>.txt:
share of samples per library ("where the time goes") and the top functions by self/total time.
"""
from __future__ import annotations
import argparse
import collections
import shutil
import subprocess
import sys
import tempfile
import time
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parent))
import run as bench # noqa: E402
ROOT = bench.ROOT
IMAGE = "mtusi-pyspy:lab3"
CATEGORIES = [ # first match wins; path fragments of the frames' files
("psycopg (PostgreSQL driver + pool)", ("psycopg",)),
("redis-py (Valkey client)", ("redis/",)),
("uvicorn + httptools (HTTP server)", ("uvicorn", "httptools")),
("starlette (routing, middleware, responses)", ("starlette",)),
("logging (access log JSON lines)", ("logging/", "logging_setup.py")),
("json (stdlib)", ("json/",)),
("asyncio / uvloop (event loop)", ("asyncio", "uvloop")),
("messenger (our code)", ("messenger/",)),
]
def category(stack: list[str]) -> str:
"""The innermost frame that belongs to a known library decides the category."""
for frame in reversed(stack):
for name, needles in CATEGORIES:
if any(n in frame for n in needles):
return name
return "other"
def short(frame: str) -> str:
name, _, where = frame.partition(" (")
file = where.rstrip(")").split(":")[0].rsplit("/", 2)[-2:]
return f"{name} ({'/'.join(file)})"
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
parser.add_argument("--label", required=True)
parser.add_argument("--config", default="replicas=2,api_cpus=1.0")
parser.add_argument("--rate", type=int, default=260)
parser.add_argument("--seconds", type=int, default=20)
parser.add_argument("--top", type=int, default=25)
parser.add_argument("--git-ref", default="HEAD")
args = parser.parse_args()
subprocess.run(["docker", "build", "-q", "-t", IMAGE, str(ROOT / "bench" / "track1" / "pyspy")], check=True,
capture_output=True)
cfg = bench.parse_config(args.config)
sha, bench.IMAGE["tag"] = bench.build_image(args.git_ref)
bench.start_stand(cfg)
work = tempfile.mkdtemp(prefix="track1-profile-")
try:
Path(work, "chats.json").write_text(__import__("json").dumps(bench.make_chats(64)), encoding="utf-8")
warmup, duration = 8, args.seconds + 8
k6 = subprocess.Popen(["docker", "compose", "-p", bench.PROJECT, *bench.FILES, "run", "--rm", "--no-deps",
"-v", f"{work}:/data:ro", "-e", f"RATE={args.rate}", "-e", f"WARMUP={warmup}s",
"-e", f"DURATION={duration}s", "-e", f"MAX_VUS={args.rate * 2}", "k6", "run", "--quiet",
"/scripts/messages.js"], cwd=ROOT, stdout=subprocess.PIPE, stderr=subprocess.DEVNULL,
text=True, encoding="utf-8")
time.sleep(warmup + 4)
target = bench.compose("ps", "-q", "api").split()[0]
subprocess.run(["docker", "run", "--rm", "--pid", f"container:{target}", "--cap-add", "SYS_PTRACE",
"-v", f"{work}:/out", IMAGE, "record", "--pid", "1", "--duration", str(args.seconds),
"--rate", "250", "--format", "raw", "--nonblocking", "--output", "/out/stacks.txt"],
check=True, capture_output=True, timeout=args.seconds + 120)
out, _ = k6.communicate(timeout=300)
data = __import__("json").loads(out[out.index(bench.MARK) + len(bench.MARK):].strip().splitlines()[0])
ok_rps = (bench.metric(data, "ok_messages{phase:measure}", "count") or 0) / duration
stacks = []
for line in Path(work, "stacks.txt").read_text(encoding="utf-8").splitlines():
stack, _, count = line.rpartition(" ")
frames = [f for f in stack.split(";") if f and not f.startswith("process ") and not f.startswith("thread ")]
if frames:
stacks.append((frames, int(count)))
finally:
shutil.rmtree(work, ignore_errors=True)
bench.compose("down", "-v", "--remove-orphans", check=False, timeout=300)
total = sum(c for _, c in stacks)
idle = sum(c for f, c in stacks if any("select" in x and ("selectors" in x or "uvloop" in x) for x in f[-2:]))
cats = collections.Counter()
self_time, total_time = collections.Counter(), collections.Counter()
for frames, count in stacks:
cats[category(frames)] += count
self_time[short(frames[-1])] += count
for name in {short(f) for f in frames}:
total_time[name] += count
lines = [f"# py-spy, {args.seconds} s at 250 Hz, ONE replica of {bench.config_name(cfg)}, k6 {args.rate} req/s "
f"(ok {ok_rps:.1f} req/s for the whole stand)", f"# label {args.label}, commit {sha}; samples {total} "
f"(idle in the event loop poll: {100 * idle / max(total, 1):.1f} %)", "",
"## Share of samples by library (innermost known frame)"]
for name, count in cats.most_common():
lines.append(f"{100 * count / total:6.1f} % {name}")
lines += ["", f"## Top {args.top} functions by SELF time (on top of the stack)"]
for name, count in self_time.most_common(args.top):
lines.append(f"{100 * count / total:6.1f} % {name}")
lines += ["", f"## Top {args.top} functions by TOTAL time (anywhere in the stack)"]
for name, count in total_time.most_common(args.top):
lines.append(f"{100 * count / total:6.1f} % {name}")
dest = ROOT / "evidence" / "track1" / f"profile-{args.label}.txt"
dest.parent.mkdir(parents=True, exist_ok=True)
dest.write_text("\n".join(lines) + "\n", encoding="utf-8")
print("\n".join(lines[:40]))
return 0
if __name__ == "__main__":
raise SystemExit(main())
+6
View File
@@ -0,0 +1,6 @@
# Profiler sidecar: shares the PID namespace of ONE api container (--pid=container:<id>) and
# gets SYS_PTRACE only for itself; the API image stays unprivileged and unchanged.
FROM python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
COPY requirements.txt /tmp/requirements.txt
RUN pip install --no-cache-dir --root-user-action=ignore --require-hashes --no-deps -r /tmp/requirements.txt
ENTRYPOINT ["py-spy"]
+3
View File
@@ -0,0 +1,3 @@
# py-spy sampling profiler for the track 1 sidecar (bench/track1/profile.py), pinned by hash.
py-spy==0.4.2 \
--hash=sha256:aeb0323409199c785f730645e9f4bb7a7b9ca2c481f2c331a55642b5d13fa52f
+396
View File
@@ -0,0 +1,396 @@
#!/usr/bin/env python3
"""Track 1 orchestrator: cost of scaling. For every configuration (replicas x CPU per replica x
PostgreSQL CPU x pool size x env) and every repeat a FRESH bench stand is started (separate
compose project, empty volumes), synthetic users/chats are created, then k6 runs an open-loop
arrival-rate sweep. During each step the orchestrator samples CPU of every container
(docker stats), DB pool waits (API stats log), PostgreSQL counters (pg_stat_database,
pg_stat_statements) and HAProxy counters (show stat).
python bench/track1/run.py --series fixed-total --label baseline \
--config replicas=1,api_cpus=2.0 --config replicas=2,api_cpus=1.0 --config replicas=4,api_cpus=0.5 \
--rates 200,400,600,800 --repeats 3
Output: evidence/track1/<series>/<label>/raw/*.json (one per step, no tokens) and
results.csv (one row per step). The main stand (project mtusi-lab3) is not touched,
but stop it during experiments so it does not compete for CPU.
"""
from __future__ import annotations
import argparse
import base64
import csv
import datetime as dt
import json
import os
import platform
import shutil
import statistics
import subprocess
import tempfile
import threading
import time
import urllib.error
import urllib.request
import uuid
from pathlib import Path
ROOT = Path(__file__).resolve().parents[2]
PROJECT = "mtusi-lab3-bench"
PORT = 18080
FILES = ["-f", str(ROOT / "compose.yaml"), "-f", str(ROOT / "bench" / "track1" / "compose.bench.yaml")]
MARK = "@@SUMMARY@@"
DEFAULTS = {"replicas": 2, "api_cpus": 1.0, "api_memory": "384m", "pg_cpus": 2.0, "pool": 10}
# ------------------------------------------------------------------ helpers
def compose(*args: str, env: dict | None = None, check: bool = True, timeout: float = 600) -> str:
result = subprocess.run(["docker", "compose", "-p", PROJECT, *FILES, *args], cwd=ROOT, capture_output=True,
text=True, encoding="utf-8", env={**os.environ, **(env or {})}, timeout=timeout)
if check and result.returncode != 0:
raise RuntimeError(f"compose {' '.join(args[:3])} failed: {result.stderr.strip()[-1500:]}")
return result.stdout
def psql(sql: str) -> str:
return compose("exec", "-T", "postgres", "psql", "-U", "messenger", "-d", "messenger", "-At", "-F", "\t",
"-v", "ON_ERROR_STOP=1", "-c", sql).strip()
def http(method: str, path: str, body=None, headers=None):
data = None if body is None else json.dumps(body).encode()
h = {"Content-Type": "application/json", **(headers or {})}
req = urllib.request.Request(f"http://127.0.0.1:{PORT}{path}", data=data, headers=h, method=method)
try:
with urllib.request.urlopen(req, timeout=30) as r:
return r.status, json.loads(r.read() or b"null")
except urllib.error.HTTPError as e:
return e.code, json.loads(e.read() or b"null")
def parse_config(text: str) -> dict:
cfg = dict(DEFAULTS)
cfg["env"] = {}
for part in filter(None, text.split(",")):
key, value = part.split("=", 1)
if key in ("replicas", "pool"):
cfg[key] = int(value)
elif key in ("api_cpus", "pg_cpus"):
cfg[key] = float(value)
elif key == "api_memory":
cfg[key] = value
else:
cfg["env"][key] = value # any API/compose env, e.g. MESSAGE_WRITE_PATH=function
return cfg
def config_name(cfg: dict) -> str:
extra = "".join(f"-{k.lower()}_{v}" for k, v in sorted(cfg["env"].items()))
return f"n{cfg['replicas']}-api{cfg['api_cpus']:g}-pg{cfg['pg_cpus']:g}-pool{cfg['pool']}{extra}"
def build_image(ref: str) -> tuple[str, str]:
"""Image of the API built from a COMMIT (git archive), not from the working tree, so a
series is reproducible and edits made while it runs cannot leak into it."""
sha = subprocess.run(["git", "rev-parse", "--short", ref], cwd=ROOT, capture_output=True, text=True,
check=True).stdout.strip()
tag = f"mtusi-messenger:bench-{sha}"
archive = subprocess.run(["git", "archive", "--format=tar", sha], cwd=ROOT, capture_output=True, check=True).stdout
subprocess.run(["docker", "build", "-q", "--build-arg", f"BUILD_REVISION={sha}", "-t", tag, "-"], input=archive,
capture_output=True, check=True, timeout=900)
return sha, tag
IMAGE = {"tag": "mtusi-messenger:lab3"}
def stand_env(cfg: dict) -> dict:
return {"API_IMAGE": IMAGE["tag"], "API_REPLICAS": str(cfg["replicas"]), "API_CPUS": f"{cfg['api_cpus']:g}", "API_MEMORY": cfg["api_memory"],
"PG_CPUS": f"{cfg['pg_cpus']:g}", "DB_POOL_MIN": str(min(cfg["pool"], 4)), "DB_POOL_MAX": str(cfg["pool"]),
"APP_PUBLISH_PORT": str(PORT), "LB_STATS_PORT": "18404", "BUILD_REVISION": "bench",
**cfg["env"]}
def start_stand(cfg: dict) -> None:
compose("down", "-v", "--remove-orphans", check=False, timeout=300)
compose("up", "-d", "--no-build", "--wait", env=stand_env(cfg), timeout=900)
psql("CREATE EXTENSION IF NOT EXISTS pg_stat_statements")
deadline = time.monotonic() + 60
while time.monotonic() < deadline: # every replica UP in HAProxy before measuring
if lb_stat()["up"] == cfg["replicas"]:
return
time.sleep(0.5)
raise RuntimeError("not all replicas UP in HAProxy")
def actual_limits() -> dict:
"""What Docker really applied (not what the config intended): CPU quota, memory, cpuset."""
ids = compose("ps", "-q").split()
result = {}
for item in json.loads(subprocess.run(["docker", "inspect", *ids], capture_output=True, text=True,
check=True).stdout):
host = item["HostConfig"]
result[item["Name"].lstrip("/")[len(PROJECT) + 1:]] = {
"cpus": host["NanoCpus"] / 1e9, "memory_mib": host["Memory"] // (1024 * 1024),
"cpuset": host["CpusetCpus"], "image": item["Config"]["Image"]}
return result
def make_chats(users: int) -> list[dict]:
created = []
stamp = uuid.uuid4().hex[:8]
for i in range(users):
username, password = f"b_{stamp}_{i}", f"Bench-password-{stamp}-{i}"
status, user = http("POST", "/api/v1/users", {"username": username, "display_name": f"bench {i}", "password": password})
assert status == 201, (status, user)
cred = base64.b64encode(f"{username}:{password}".encode()).decode()
status, session = http("POST", "/api/v1/auth/sessions", headers={"Authorization": "Basic " + cred})
assert status == 201, (status, session)
created.append({"id": user["id"], "token": session["token"]})
chats = []
for i, me in enumerate(created):
other = created[(i + 1) % len(created)]
status, chat = http("POST", "/api/v1/chats", {"title": f"bench {i}", "member_ids": [me["id"], other["id"]]},
headers={"Authorization": "Bearer " + me["token"]})
assert status == 201, (status, chat)
chats.append({"token": me["token"], "chat": chat["id"]})
return chats
class Sampler(threading.Thread):
"""docker stats every ~1-2 s: CPU% (100 = one core) and memory per container."""
def __init__(self):
super().__init__(daemon=True)
self.samples: list[tuple[float, dict]] = []
self.stop_event = threading.Event()
def run(self) -> None:
while not self.stop_event.is_set():
out = subprocess.run(["docker", "stats", "--no-stream", "--format", "{{json .}}"],
capture_output=True, text=True).stdout
now = time.monotonic()
row = {}
for line in out.splitlines():
item = json.loads(line)
if not item["Name"].startswith(PROJECT + "-"):
continue # only this bench project (k6 run containers included)
short = item["Name"][len(PROJECT) + 1:]
if short.startswith("k6-run-"):
short = "k6"
row[short] = {"cpu": float(item["CPUPerc"].rstrip("%") or 0), "mem": item["MemUsage"].split(" / ")[0]}
self.samples.append((now, row))
def window(self, start: float, end: float) -> dict:
rows = [r for t, r in self.samples if start <= t <= end]
result = {}
# All API replicas summed PER SAMPLE first, then the median over time (not a sum of medians).
totals = [sum(v["cpu"] for k, v in r.items() if k.startswith("api-")) for r in rows]
if totals:
result["api_total"] = {"cpu_median": round(statistics.median(totals), 1), "cpu_max": round(max(totals), 1),
"samples": len(totals)}
for name in sorted({n for r in rows for n in r}):
values = [r[name]["cpu"] for r in rows if name in r]
result[name] = {"cpu_median": round(statistics.median(values), 1), "cpu_max": round(max(values), 1),
"mem_last": [r[name]["mem"] for r in rows if name in r][-1], "samples": len(values)}
return result
def lb_stat() -> dict:
raw = compose("exec", "-T", "lb", "sh", "-c", 'echo "show stat" | socat stdio /var/lib/haproxy/admin.sock')
lines = [line for line in raw.splitlines() if line]
header = lines[0].lstrip("# ").split(",")
servers = [dict(zip(header, line.split(","))) for line in lines[1:]]
api = [s for s in servers if s["pxname"] == "api" and s["svname"].startswith("api") and s["status"] != "MAINT"]
backend = next(s for s in servers if s["pxname"] == "api" and s["svname"] == "BACKEND")
return {"up": sum(1 for s in api if s["status"].startswith("UP")),
"chkdown": sum(int(s["chkdown"] or 0) for s in api), "qmax": int(backend["qmax"] or 0),
"smax": int(backend["smax"] or 0), "econ": int(backend["econ"] or 0), "eresp": int(backend["eresp"] or 0),
"wredis": int(backend["wredis"] or 0), "wretr": int(backend["wretr"] or 0)}
def pg_counters() -> dict:
row = psql("SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database "
"WHERE datname = 'messenger'").split("\t")
wal = psql("SELECT wal_records, wal_bytes FROM pg_stat_wal").split("\t")
return {"xact_commit": int(row[0]), "xact_rollback": int(row[1]), "tup_inserted": int(row[2]),
"blks_read": int(row[3]), "blks_hit": int(row[4]), "wal_records": int(wal[0]), "wal_bytes": int(float(wal[1]))}
def top_statements(limit: int = 12) -> list[dict]:
out = psql("SELECT calls, round(total_exec_time::numeric, 1), round(mean_exec_time::numeric, 4), "
"left(regexp_replace(query, '\\s+', ' ', 'g'), 110) FROM pg_stat_statements "
"WHERE dbid = (SELECT oid FROM pg_database WHERE datname = 'messenger') "
f"ORDER BY total_exec_time DESC LIMIT {limit}")
rows = []
for line in out.splitlines():
calls, total, mean, query = line.split("\t", 3)
rows.append({"calls": int(calls), "total_ms": float(total), "mean_ms": float(mean), "query": query})
return rows
def pool_window(since_iso: str, start_wall: float, end_wall: float) -> dict:
logs = compose("logs", "--no-log-prefix", "--since", since_iso, "api")
waits, nums, waiting, cpu, usage, intervals = 0, 0, 0, 0.0, 0, 0
for line in logs.splitlines():
if '"pool_stats"' not in line:
continue
item = json.loads(line)
ts = dt.datetime.fromisoformat(item["ts"].replace("Z", "+00:00")).timestamp()
if not start_wall <= ts <= end_wall:
continue
intervals += 1
waits += item["requests_wait_ms"]
nums += item["requests_num"]
waiting = max(waiting, item["requests_waiting"])
cpu += item["cpu_s"]
usage += item["usage_ms"]
return {"pool_requests": nums, "pool_wait_ms_avg": round(waits / nums, 3) if nums else None,
"pool_waiting_max": waiting, "pool_usage_ms_avg": round(usage / nums, 3) if nums else None,
"api_cpu_s": round(cpu, 2), "intervals": intervals}
def metric(data: dict, name: str, field: str):
m = data["metrics"].get(name)
return None if m is None else m["values"].get(field)
def run_step(cfg: dict, rate: int, args, chats_dir: str, sampler: Sampler) -> dict:
psql("SELECT pg_stat_statements_reset()")
pg0, lb0 = pg_counters(), lb_stat()
since = dt.datetime.now(dt.timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
started = time.monotonic()
wall0 = time.time()
max_vus = max(50, min(args.max_vus, rate * 2))
out = compose("run", "--rm", "--no-deps", "-v", f"{chats_dir}:/data:ro", "-e", f"RATE={rate}",
"-e", f"WARMUP={args.warmup}s", "-e", f"DURATION={args.duration}s", "-e", f"MAX_VUS={max_vus}",
"k6", "run", "--quiet", "/scripts/messages.js", timeout=args.warmup + args.duration + 180)
ended = time.monotonic()
data = json.loads(out[out.index(MARK) + len(MARK):].strip().splitlines()[0])
# k6 starts ~1-3 s after the command; the measure phase is the last `duration` seconds before exit.
m_end = ended - 1.0
m_start = max(started, m_end - args.duration)
pg1, lb1 = pg_counters(), lb_stat()
ok = metric(data, "ok_messages{phase:measure}", "count") or 0
reqs = metric(data, "http_reqs{phase:measure}", "count") or 0
dropped = metric(data, "dropped_iterations{scenario:measure}", "count") or 0
result = {
"config": config_name(cfg), "rate": rate, "duration_s": args.duration,
"offered": rate * args.duration, "requests": int(reqs), "ok": int(ok), "dropped": int(dropped),
"ok_rps": round(ok / args.duration, 1),
"error_ratio": round(metric(data, "http_req_failed{phase:measure}", "rate") or 0, 5),
"ok_p50_ms": metric(data, "ok_latency{phase:measure}", "p(50)"),
"ok_p95_ms": metric(data, "ok_latency{phase:measure}", "p(95)"),
"ok_p99_ms": metric(data, "ok_latency{phase:measure}", "p(99)"),
"vus_max": metric(data, "vus_max", "value"),
"cpu": sampler.window(m_start, m_end),
"pool": pool_window(since, wall0 + (m_start - started), wall0 + (m_end - started)),
"pg": {k: pg1[k] - pg0[k] for k in pg1},
"lb": {"chkdown_delta": lb1["chkdown"] - lb0["chkdown"], "qmax": lb1["qmax"], "up_after": lb1["up"],
"econ_delta": lb1["econ"] - lb0["econ"], "eresp_delta": lb1["eresp"] - lb0["eresp"],
"redispatch_delta": lb1["wredis"] - lb0["wredis"]},
"statements": top_statements(),
}
commits = result["pg"]["xact_commit"]
result["pg_commits_per_ok"] = round(commits / ok, 2) if ok else None
result["statements_per_ok"] = round(sum(s["calls"] for s in result["statements"]) / ok, 2) if ok else None
return result
def flat(r: dict) -> dict:
cpu = r["cpu"]
api_cpu = cpu.get("api_total", {}).get("cpu_median", 0.0)
limits = r.get("limits", {})
api_limits = [v for k, v in limits.items() if k.startswith("api-")]
return {"config": r["config"], "repeat": r["repeat"], "rate": r["rate"], "ok_rps": r["ok_rps"],
"replicas_running": len(api_limits), "api_cpus_limit_total": round(sum(v["cpus"] for v in api_limits), 2),
"api_mem_mib_each": api_limits[0]["memory_mib"] if api_limits else None,
"pg_cpus_limit": limits.get("postgres-1", {}).get("cpus"),
"requests": r["requests"], "ok": r["ok"], "dropped": r["dropped"], "error_ratio": r["error_ratio"],
"p50_ms": r["ok_p50_ms"], "p95_ms": r["ok_p95_ms"], "p99_ms": r["ok_p99_ms"],
"api_cpu_pct_total": round(api_cpu, 1), "pg_cpu_pct": cpu.get("postgres-1", {}).get("cpu_median"),
"lb_cpu_pct": cpu.get("lb-1", {}).get("cpu_median"), "valkey_cpu_pct": cpu.get("valkey-1", {}).get("cpu_median"),
"k6_cpu_pct": cpu.get("k6", {}).get("cpu_median"),
"pool_wait_ms_avg": r["pool"]["pool_wait_ms_avg"], "pool_waiting_max": r["pool"]["pool_waiting_max"],
"pg_commits_per_ok": r["pg_commits_per_ok"], "statements_per_ok": r["statements_per_ok"],
"lb_chkdown": r["lb"]["chkdown_delta"], "lb_qmax": r["lb"]["qmax"], "wal_bytes": r["pg"]["wal_bytes"]}
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])
parser.add_argument("--series", required=True)
parser.add_argument("--label", default="baseline")
parser.add_argument("--config", action="append", required=True, help="replicas=2,api_cpus=1.0,pg_cpus=2.0,pool=10,ENV=val")
parser.add_argument("--rates", required=True, help="comma separated arrival rates, req/s")
parser.add_argument("--repeats", type=int, default=3)
parser.add_argument("--warmup", type=int, default=10)
parser.add_argument("--duration", type=int, default=20)
parser.add_argument("--users", type=int, default=64)
parser.add_argument("--max-vus", type=int, default=600)
parser.add_argument("--git-ref", default="HEAD", help="commit whose code is measured (image built via git archive)")
args = parser.parse_args()
sha, IMAGE["tag"] = build_image(args.git_ref)
out = ROOT / "evidence" / "track1" / args.series / args.label
(out / "raw").mkdir(parents=True, exist_ok=True)
rates = [int(r) for r in args.rates.split(",")]
configs = [parse_config(c) for c in args.config]
meta = {"started": dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"),
"measured_commit": sha, "image": IMAGE["tag"],
"bench_tooling_dirty": bool(subprocess.run(["git", "status", "--porcelain", "--", "bench", "compose.yaml",
"infra"], cwd=ROOT, capture_output=True, text=True).stdout.strip()),
"host": {"os": f"{platform.system()} {platform.release()}",
"docker": subprocess.run(["docker", "info", "--format", "{{.ServerVersion}} {{.NCPU}} CPU "
"{{.MemTotal}} B {{.KernelVersion}}"], capture_output=True,
text=True).stdout.strip()},
"cpusets": "k6 0-1, lb+valkey 2, postgres 3-4, api 5-7 (bench/track1/compose.bench.yaml)",
"workload": f"k6 constant-arrival-rate POST message, new Idempotency-Key, {args.users} chats, "
f"warmup {args.warmup}s + measure {args.duration}s per step, fresh stand per config x repeat",
"rates": rates, "repeats": args.repeats, "configs": configs, "args": sys_argv()}
(out / "meta.json").write_text(json.dumps(meta, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
csv_path = out / "results.csv"
new_file = not csv_path.exists()
with csv_path.open("a", newline="", encoding="utf-8") as f:
writer = None
for repeat in range(1, args.repeats + 1):
for cfg in configs:
name = config_name(cfg)
print(f"== {args.series}/{args.label} {name} repeat {repeat}: fresh stand", flush=True)
start_stand(cfg)
limits = actual_limits()
chats_dir = tempfile.mkdtemp(prefix="track1-")
try:
Path(chats_dir, "chats.json").write_text(json.dumps(make_chats(args.users)), encoding="utf-8")
sampler = Sampler()
sampler.start()
for rate in rates:
result = run_step(cfg, rate, args, chats_dir, sampler)
result["repeat"] = repeat
result["limits"] = limits
(out / "raw" / f"{name}-r{repeat}-{rate}.json").write_text(
json.dumps(result, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
row = flat(result)
if writer is None:
writer = csv.DictWriter(f, fieldnames=list(row))
if new_file:
writer.writeheader()
writer.writerow(row)
f.flush()
print(f" rate {rate}: ok_rps {row['ok_rps']} p95 {row['p95_ms'] and round(row['p95_ms'])} ms "
f"err {row['error_ratio']} dropped {row['dropped']} | api {row['api_cpu_pct_total']}% "
f"pg {row['pg_cpu_pct']}% lb {row['lb_cpu_pct']}% k6 {row['k6_cpu_pct']}% "
f"pool_wait {row['pool_wait_ms_avg']} ms chkdown {row['lb_chkdown']}", flush=True)
sampler.stop_event.set()
finally:
shutil.rmtree(chats_dir, ignore_errors=True) # synthetic tokens never stay on disk
compose("down", "-v", "--remove-orphans", check=False, timeout=300)
print(f"results: {csv_path}")
return 0
def sys_argv() -> list[str]:
import sys
return sys.argv[1:]
if __name__ == "__main__":
raise SystemExit(main())
+44
View File
@@ -0,0 +1,44 @@
#!/bin/sh
# The exact track 1 experiment plan (REPORT.md, "Трек 1"). Every series starts fresh stands and
# measures the API image built from the given commit. Stop the main stand first (CPU noise):
# docker compose stop
# sh bench/track1/series.sh baseline <commit-before-the-improvement>
# sh bench/track1/series.sh improved <commit-with-the-improvement>
# sh bench/track1/series.sh ceiling <commit-with-the-improvement>
# ~2.5 h for `baseline`, ~1.5 h for `improved` on the reference laptop (see meta.json files).
set -eu
cd "$(dirname "$0")/../.."
PY="${PYTHON:-$(command -v python3 || command -v python)}"
phase="${1:?baseline, improved or ceiling}"
ref="${2:-HEAD}"
run() { "$PY" bench/track1/run.py --git-ref "$ref" --repeats "${REPEATS:-3}" "$@"; }
case "$phase" in
baseline)
# A. Fixed total API budget 2.0 CPU / 768 MiB split into 1, 2, 4 replicas; PostgreSQL 2.0 CPU.
run --series A-fixed-total --label baseline --rates 100,150,200,250,300,350 \
--config replicas=1,api_cpus=2.0,api_memory=768m \
--config replicas=2,api_cpus=1.0,api_memory=384m \
--config replicas=4,api_cpus=0.5,api_memory=192m
# B. Growing API resources (1, 2, 3 replicas x 1.0 CPU) against a NARROW database (0.5 CPU).
run --series B-replicas-vs-db --label baseline --rates 100,150,200,250,300,350,400 \
--config replicas=1,pg_cpus=0.5 --config replicas=2,pg_cpus=0.5 --config replicas=3,pg_cpus=0.5
# C. Growing database resources (0.5 -> 1.0 -> 2.0 CPU) at 3 API replicas: does the ceiling move?
run --series C-db-growth --label baseline --rates 200,250,300,350,400,450,500 \
--config replicas=3,pg_cpus=1.0 --config replicas=3,pg_cpus=2.0
;;
improved)
# D. The improvement (single-round-trip write) against the old write path on the SAME image,
# in the API-bound (2 x 1.0, PG 2.0) and the DB-bound (3 x 1.0, PG 0.5) configurations.
run --series D-improvement --label "$ref" --rates 150,200,250,300,350,400,450,500 \
--config replicas=2,MESSAGE_WRITE_PATH=multi --config replicas=2,MESSAGE_WRITE_PATH=function \
--config replicas=3,pg_cpus=0.5,MESSAGE_WRITE_PATH=multi --config replicas=3,pg_cpus=0.5,MESSAGE_WRITE_PATH=function
;;
ceiling)
# E. Series D showed the function path NOT saturated at 500/s with 2 x 1.0: find its ceiling
# (same configurations, higher rates) and where the bottleneck moves to.
run --series E-improved-ceiling --label "$ref" --rates 450,550,650,750,850 \
--config replicas=2,MESSAGE_WRITE_PATH=function --config replicas=3,pg_cpus=0.5,MESSAGE_WRITE_PATH=function
;;
*) echo "usage: series.sh baseline|improved|ceiling <git-ref>" >&2; exit 2 ;;
esac
+184
View File
@@ -0,0 +1,184 @@
# Lab 3 stand: client -> HAProxy (lb) -> API x N (one scalable service `api`) -> PostgreSQL + Valkey.
# From a clean clone (details and Windows notes: REPORT.md, docs/WINDOWS.md):
# python scripts/init_secrets.py # once: random local credentials into ./secrets (git-ignored)
# cp .env.example .env # optional overrides (API_REPLICAS, limits, ...)
# docker compose --env-file .env up -d --build --wait
# Order: postgres/valkey healthy -> migrate (ONE run, exits 0) -> api replicas healthy -> lb.
# Images are pinned by version AND digest. Passwords are passed only as secret files, and
# every service gets only the secrets it needs. Only the LB publishes ports (loopback only).
# Two networks: `edge` (lb <-> api) and `data` (api/migrate <-> postgres/valkey, internal):
# the LB cannot reach the stores, and the stores are not reachable from the edge.
name: mtusi-lab3
x-api-env: &api-env
APP_ENV: ${APP_ENV:-lab}
LOG_LEVEL: ${LOG_LEVEL:-info}
SESSION_TTL_SECONDS: ${SESSION_TTL_SECONDS:-3600}
SESSION_STORE: ${SESSION_STORE:-valkey}
DATABASE_URL_FILE: /run/secrets/database_url
CACHE_URL_FILE: /run/secrets/cache_url
DB_POOL_MIN: ${DB_POOL_MIN:-1}
DB_POOL_MAX: ${DB_POOL_MAX:-10}
DB_TIMEOUT_SECONDS: ${DB_TIMEOUT_SECONDS:-3}
CACHE_TIMEOUT_SECONDS: ${CACHE_TIMEOUT_SECONDS:-0.5}
PASSWORD_HASH_CONCURRENCY: ${PASSWORD_HASH_CONCURRENCY:-2}
IDEMPOTENCY_TTL_SECONDS: ${IDEMPOTENCY_TTL_SECONDS:-86400}
# Graceful stop: 3 s drain (readiness 503, still serving) + up to 8 s for in-flight requests.
DRAIN_SECONDS: ${DRAIN_SECONDS:-3}
SHUTDOWN_TIMEOUT_SECONDS: ${SHUTDOWN_TIMEOUT_SECONDS:-8}
DOCS_ENABLED: ${DOCS_ENABLED:-true}
# Track 1 improvement: `function` = one SQL call per send (default), `multi` = old path (A/B).
MESSAGE_WRITE_PATH: ${MESSAGE_WRITE_PATH:-function}
services:
postgres:
image: postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c
environment:
POSTGRES_USER: messenger
POSTGRES_DB: messenger
POSTGRES_PASSWORD_FILE: /run/secrets/postgres_password
# pg_stat_statements: per-statement counters for the track 5 experiments (bench/track1).
command: ["postgres", "-c", "shared_preload_libraries=pg_stat_statements", "-c", "max_connections=${PG_MAX_CONNECTIONS:-100}"]
secrets: [postgres_password]
volumes:
- pgdata:/var/lib/postgresql # PostgreSQL 18 image keeps PGDATA under this path
healthcheck:
test: ["CMD-SHELL", "pg_isready -U messenger -d messenger -q"]
interval: 2s
timeout: 3s
retries: 30
start_period: 10s
stop_grace_period: 30s # clean shutdown checkpoint instead of crash recovery
restart: unless-stopped
networks: [data]
deploy:
resources:
limits: {cpus: "${PG_CPUS:-2.0}", memory: "${PG_MEMORY:-1g}"}
valkey:
image: valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571
# The password goes into a config include readable only by the valkey user,
# so it never appears in the process arguments. docker-entrypoint.sh drops root.
command:
- sh
- -c
- >-
umask 077 &&
printf 'requirepass %s\n' "$$(cat /run/secrets/valkey_password)" > /tmp/requirepass.conf &&
chown valkey /tmp/requirepass.conf &&
exec docker-entrypoint.sh valkey-server /etc/valkey/valkey.conf
secrets: [valkey_password]
volumes:
- ./infra/valkey/valkey.conf:/etc/valkey/valkey.conf:ro
- valkeydata:/data
healthcheck:
test: ["CMD-SHELL", "REDISCLI_AUTH=\"$$(cat /run/secrets/valkey_password)\" valkey-cli ping | grep -q PONG"]
interval: 2s
timeout: 3s
retries: 30
start_period: 5s
restart: unless-stopped
networks: [data]
deploy:
resources:
limits: {cpus: "${VALKEY_CPUS:-0.5}", memory: 256m}
# One-shot schema migration: runs ONCE per `up`, before any API replica starts
# (api depends_on service_completed_successfully). Replicas never migrate themselves.
migrate:
build:
context: .
args:
BUILD_REVISION: ${BUILD_REVISION:-dev}
image: ${API_IMAGE:-mtusi-messenger:lab3}
command: ["python", "-m", "messenger.migrate"]
environment:
DATABASE_URL_FILE: /run/secrets/database_url
secrets: [database_url]
depends_on:
postgres:
condition: service_healthy
healthcheck:
disable: true
restart: "no"
networks: [data]
deploy:
resources:
limits: {cpus: "0.5", memory: 256m}
api:
build:
context: .
args:
BUILD_REVISION: ${BUILD_REVISION:-dev}
image: ${API_IMAGE:-mtusi-messenger:lab3}
environment: *api-env
secrets: [database_url, cache_url]
# No `ports` and no container_name: replicas are reachable only through `lb`,
# and `--scale api=N` / deploy.replicas works. X-Instance-Id = container hostname.
expose: ["8080"]
healthcheck:
test: ["CMD", "python", "-m", "messenger.healthcheck"]
interval: 2s
timeout: 2s
retries: 3
start_period: 5s
depends_on:
postgres:
condition: service_healthy
valkey:
condition: service_healthy
migrate:
condition: service_completed_successfully
stop_grace_period: 15s # > DRAIN_SECONDS + SHUTDOWN_TIMEOUT_SECONDS (11 s)
restart: unless-stopped
networks: [edge, data]
deploy:
replicas: ${API_REPLICAS:-2}
resources:
limits: {cpus: "${API_CPUS:-1.0}", memory: "${API_MEMORY:-256m}"}
lb:
image: haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122
volumes:
- ./infra/haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro
ports:
# The ONLY public entry point, on the loopback interfaces only. Both IPv4 and IPv6:
# on Windows `localhost` resolves to ::1 first, and a refused ::1 connection costs ~2 s
# per request before the fallback to 127.0.0.1 (measured: 2.0 s vs 0.007 s).
- "127.0.0.1:${APP_PUBLISH_PORT:-8080}:8080"
- "[::1]:${APP_PUBLISH_PORT:-8080}:8080"
- "127.0.0.1:${LB_STATS_PORT:-8404}:8404" # local status page (UP/DOWN of replicas)
healthcheck:
# Config is valid and the process answers; whether any API is ready is the LB's own job.
test: ["CMD-SHELL", "haproxy -c -q -f /usr/local/etc/haproxy/haproxy.cfg && wget -q -O /dev/null http://127.0.0.1:8404/"]
interval: 5s
timeout: 3s
retries: 3
depends_on:
api:
condition: service_healthy
restart: unless-stopped
networks: [edge]
deploy:
resources:
limits: {cpus: "${LB_CPUS:-1.0}", memory: 128m}
networks:
edge:
data:
internal: true # no route outside: stores talk only to api/migrate
volumes:
pgdata:
valkeydata:
secrets:
postgres_password:
file: ./secrets/postgres_password
valkey_password:
file: ./secrets/valkey_password
database_url:
file: ./secrets/database_url
cache_url:
file: ./secrets/cache_url
+43
View File
@@ -0,0 +1,43 @@
# Модель данных lab3
Схема: [0001_init.sql](../migrations/0001_init.sql) (lab2), [0002_idempotency_keys.sql](../migrations/0002_idempotency_keys.sql) и [0003_post_message_function.sql](../migrations/0003_post_message_function.sql) (lab3). Миграции применяет один раз отдельный сервис `migrate` до старта реплик API; обе миграции lab3 только добавляют объекты, поэтому совместимы с работающей предыдущей версией (rolling update). Источник истины для пользователей, чатов, участников и сообщений — PostgreSQL 18. Для сессий источник истины — хранилище, выбранное через `SESSION_STORE`. По умолчанию это Valkey с AOF, его можно переключить на таблицу `sessions` в PostgreSQL.
## Таблицы и ограничения
| Таблица | Ключ | Ограничения | Зачем |
| --- | --- | --- | --- |
| `users` | `id uuid` | `UNIQUE (username)`, `CHECK` формата username, длина `display_name` | уникальность имени держит БД, а не код: гонку регистрации разрешает уникальный индекс |
| `chats` | `id uuid` | FK `created_by → users`, длина `title` | `last_message_at` — «часы» чата для упорядочивания сообщений |
| `chat_members` | `(chat_id, user_id)` | FK на `chats` и `users`, `UNIQUE (chat_id, position)` | `position` хранит порядок `member_ids`; `chat_created_at` — копия времени чата для keyset-списка чатов |
| `messages` | `id uuid` | FK `(chat_id, sender_id) → chat_members`, длина `text` | отправитель обязан быть участником — это гарантирует сама БД |
| `sessions` | `token_hash bytea` | FK `user_id → users`, длина хеша 32 байта | хранится SHA-256 токена; сам токен не хранится нигде |
| `idempotency_keys` | `(user_id, chat_id, key)` | FK на `users`, `chats`; `message_id → messages` (DEFERRABLE, `UNIQUE`); `CHECK` формата ключа и длины SHA-256 | общая для всех реплик идемпотентность отправки: ключ → отпечаток тела → сообщение. Хранится ≥ `IDEMPOTENCY_TTL_SECONDS` (24 ч), очистка — пачками раз в 10 мин под advisory-замком |
| `schema_migrations` | `version` | checksum файла | какие миграции применены; изменённую задним числом миграцию мигратор не применит |
**Почему `bytea` для текста.** Контракт разрешает символ U+0000, а тип `text` в PostgreSQL не может содержать байт NUL (`DataError`). Поэтому `display_name`, `title` и `text` хранятся как байты UTF-8. Длину в code points проверяет API, а `CHECK` в байтах — страховка (1 code point ≤ 4 байта).
## Индексы и запросы
- `messages_chat_keyset (chat_id, created_at, id)`. Первая страница — `WHERE chat_id=$1 ORDER BY created_at, id LIMIT n`, следующая — `WHERE chat_id=$1 AND (created_at, id) > ($2, $3) ORDER BY ... LIMIT n`. Сравнение строк `(created_at, id) > (...)` попадает в `Index Cond`: поиск начинается прямо с позиции cursor и читает только `n+1` строк. Запросов два, а не один с `($2 IS NULL OR ...)`: иначе при generic plan условие cursor уходит в `Filter`, и поздняя страница сканирует всю историю чата до неё.
- `chat_members_user_keyset (user_id, chat_created_at, chat_id)` — список чатов пользователя тем же keyset-способом.
- `users_username_key` — вход по username и уникальность.
- `sessions_expires_at` — очистка истёкших сессий.
## Транзакционные границы
| Операция | Транзакция | Что гарантирует |
| --- | --- | --- |
| Регистрация | один `INSERT` | `UniqueViolation` → 409; ровно одна строка при гонке |
| Создание чата | `INSERT chats` → `INSERT chat_members` (все участники одним запросом) → `COMMIT` | неизвестный участник вызывает FK-ошибку **после** вставки чата; откат удаляет и чат, и частичное членство |
| Сообщение (lab3, с ключом) | проверка членства → `INSERT idempotency_keys … ON CONFLICT DO UPDATE … WHERE ключ старше 24 ч` → если вставлен: `UPDATE chats … RETURNING` → `INSERT messages`; если нет: прочитать сохранённое сообщение и сравнить отпечаток → `COMMIT`. По умолчанию всё это — одна SQL-функция `post_message` в autocommit (одно обращение к БД) | ключ, сообщение и «результат» коммитятся вместе или не коммитятся вовсе: «висящего» ключа нет. Конкурентный повтор ждёт на уникальном индексе завершения первой транзакции и затем видит её результат (READ COMMITTED, новый снимок на каждый оператор). Другое тело → 409, тот же — тот же Message |
| Сообщение (lab2-часть) | проверка членства → `UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + 1 мкс) RETURNING` → `INSERT messages` → `COMMIT` | строка чата заблокирована до конца транзакции, поэтому писатели одного чата идут по очереди. Время сообщений строго растёт **в порядке коммитов**: страница по cursor не пропустит сообщение, закоммиченное «задним числом» |
| Вход | `SELECT` пользователя, проверка Argon2id, запись сессии | неверный пароль ничего не пишет |
| Logout | `DELETE` сессии (или `DEL` в Valkey) | повтор с тем же токеном — 401 |
Чтение истории не требует snapshot между HTTP-запросами (так в контракте): новое сообщение появится на следующей странице. Список чатов сортируется по `chats.created_at = clock_timestamp()` без блокировки. Два чата, создаваемые одновременно, могут закоммититься не в порядке времени. Тогда у обхода, который уже начался, такой чат может оказаться «до» cursor; новый обход его увидит. Чаты создаются редко, поэтому мы не сериализуем их глобальным замком; это ограничение описано в REPORT.md.
## Сессии
- Токен: `secrets.token_urlsafe(32)` — 256 бит случайности, в ответе 43 символа. В хранилище лежит только SHA-256 токена. Для случайного 256-битного значения быстрого хеша достаточно, медленный KDF нужен только для низкоэнтропийных паролей.
- `SESSION_STORE=valkey` (по умолчанию): ключ `session:<sha256-hex>`, значение `{user_id, expires_at}`. Срок жизни задаётся абсолютным `PXAT expires_at`, поэтому рестарт его не продлевает. Включён `appendonly yes` + `appendfsync always`: подтверждённые вход и logout сохраняются до ответа клиенту. Valkey — **единственное** хранилище сессий, это не кэш над PostgreSQL. Поэтому нет второй копии, которая могла бы «оживить» отозванный токен после отказа. Если Valkey недоступен, проверить сессию нельзя: защищённые запросы и logout получают 503, анонимный доступ не открывается.
- `SESSION_STORE=postgres`: таблица `sessions`, проверка `expires_at > now()` в запросе.
+104
View File
@@ -0,0 +1,104 @@
# Шпаргалка к защите ЛР 3
Отчёт с доказательствами — [REPORT.md](../REPORT.md), команды на ноутбуке — [WINDOWS.md](WINDOWS.md), модель данных — [DATA_MODEL.md](DATA_MODEL.md).
## Что изменилось по сравнению с lab2
| Было в lab2 | Стало в lab3 | Где |
| --- | --- | --- |
| один контейнер `api`, порт 8080 наружу | сервис `api` × N (`deploy.replicas`, без `ports` и `container_name`), вход только через HAProxy | [compose.yaml](../compose.yaml), [haproxy.cfg](../infra/haproxy/haproxy.cfg) |
| — | `X-Instance-Id` = hostname контейнера (короткий container ID) на **каждом** ответе, включая 500/503 | `RequestContextMiddleware` в [app.py](../src/messenger/app.py) |
| POST сообщения без ключа | `Idempotency-Key` обязателен; ключ, сообщение и результат — одна транзакция | [repo.py](../src/messenger/repo.py), [0002](../migrations/0002_idempotency_keys.sql), [0003](../migrations/0003_post_message_function.sql) |
| SIGTERM → сразу закрыть сокет | SIGTERM → readiness 503 и 3 с drain (запросы ещё обслуживаются) → graceful stop | [__main__.py](../src/messenger/__main__.py) |
| миграции — сервис `migrate` | так же один раз до старта реплик; реплики сами не мигрируют | `migrate` в compose.yaml |
| одна сеть | `edge` (lb↔api) и internal `data` (api/migrate↔postgres/valkey) | compose.yaml |
| дедлайн БД отменял задачу psycopg | задача не отменяется (cancel в libpq блокировал event loop), мёртвые сокеты рвёт `tcp_user_timeout` | [db.py](../src/messenger/db.py) |
Осталось из lab2: сессии только в Valkey (AOF `always`), Argon2id, cursor с привязкой к ресурсу, строгая валидация, `bytea` для текста с U+0000, дедлайны БД/Valkey, 503 вместо 401 при отказе хранилищ.
## Схема
```
клиент ──127.0.0.1:8080──▶ HAProxy (lb) ──edge──▶ api-1, api-2 … (Starlette/uvicorn, без состояния)
│ check GET /health/ready каждую 1 с │
│ fall 2 / rise 2, Docker DNS каждые 1 с └──data──▶ PostgreSQL 18 (источник истины)
└─ свои 502/503/504 в формате ошибки контракта Valkey 8 (сессии, AOF always)
```
## Путь `POST /chats/{id}/messages` с `Idempotency-Key`
1. HAProxy: ставит `X-Request-Id` (если у клиента нет валидного), выбирает следующую `UP`-реплику (roundrobin).
2. API: `parse_bearer` → Valkey `session:<sha256>` (общий для всех реплик) → 401/503.
3. UUID чата → 400. Посторонний → **404 всегда**, даже при повторе чужого ключа (членство проверяется до ключа).
4. Ключ `[A-Za-z0-9._:-]{1,128}` и ровно один заголовок → иначе 400. Тело: строгий JSON (дубли ключей, лишние поля — 400), `attachment_ids: []` = отсутствие поля.
5. Отпечаток тела = SHA-256 канонического JSON (`sort_keys`, `attachment_ids` по умолчанию `[]`).
6. Запись — **одна транзакция** (функция `post_message`, миграция 0003, или прежний многооператорный путь `MESSAGE_WRITE_PATH=multi`):
- `INSERT INTO idempotency_keys … ON CONFLICT DO UPDATE … WHERE ключ старше 24 ч`;
- вставили ключ → `UPDATE chats` (часы чата под блокировкой строки) → `INSERT messages` → `created`;
- не вставили → ключ уже есть: тот же отпечаток → **тот же Message, 201** (`replayed`), другой → **409**.
7. Конкурентный повтор с тем же ключом через другую реплику ждёт на уникальном индексе, пока первая транзакция не завершится: после COMMIT он видит результат, после ROLLBACK вставляет сам. Отдельного «pending»-состояния нет, поэтому сбой не оставляет «висящий» ключ (это был главный блокер у других).
## Вопросы из README
**Чем отказ процесса отличается от отказа узла?**
- **Процесс** (у нас — API-реплика): остальные процессы, сеть, диск и Docker живы. Его можно перезапустить рядом, а LB перестаёт слать на него запросы: соединение отклонено → HAProxy сразу повторяет на другой (`retry-on conn-failure` + `redispatch`), readiness не отвечает → через ≈ 1,5–3,5 с `DOWN`. Данные в PostgreSQL/Valkey не страдают.
- **Узел** (хост/VM): пропадает всё, что на нём: все реплики, LB, БД, volumes и сам Docker. Пережить это можно только если копии каждого нужного компонента живут на **других** узлах: реплики API, второй LB (VIP/DNS), реплицированная БД с переключением, хранилище не на этом диске.
- У нас один Docker-host: стенд выдерживает отказ **одной API-реплики** (доказано `make test-failover` и серией в REPORT), но не отказ узла. LB, PostgreSQL и Valkey — единичные точки отказа (SPOF), это записано в ограничениях.
**Почему `depends_on`/`healthcheck` сами по себе не являются балансировщиком?**
- `depends_on` действует **только при старте**: в каком порядке запускать и чего ждать. После запуска он ничего не делает.
- `healthcheck` Docker только **помечает** контейнер `unhealthy`. Трафик он не перенаправляет и контейнер не перезапускает (в Compose без Swarm). Пометка появляется через `interval × retries`.
- Распределение запросов, исключение неготовой реплики и повтор на другую — работа LB. У нас HAProxy сам опрашивает `/health/ready` каждую секунду и не шлёт бизнес-запросы реплике, которая жива (`/health/live` 200), но не готова (нет БД или Valkey, идёт drain). Проверено тестом: у одной реплики отнят PostgreSQL (`pg_hba reject`) или вся сеть `data` (`docker network disconnect`) — через 3–4 с ни одного запроса к ней, вторая обслуживает всё (`test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready`). Резервных серверов «без проверки» нет.
**Какие POST можно повторять после timeout и что защищает `Idempotency-Key`?**
- Timeout = **неизвестный исход**: запрос мог не дойти, мог выполниться без ответа, мог выполниться и ответ потерялся.
- Повторять без вреда можно то, что идемпотентно: **отправку сообщения с тем же ключом и тем же телом**. Сервер вернёт исходный 201 и тот же Message, дубля не будет. Это проверено и для повтора через другую реплику, и для 24 одновременных повторов, и для случая, когда клиент сдался через 0,4 с, а сервер закоммитил позже.
- Регистрацию, создание чата, вход повторять «прозрачно» нельзя: без ключа сервер не отличит повтор от нового запроса (второй чат, `409` на username). Поэтому HAProxy повторяет **только** запросы, которые не были отправлены на реплику (`retry-on conn-failure`); всё, что могло дойти до API, повторяет клиент сам и осознанно.
- Ключ защищает от **дубля при повторе**, но не от другого тела с тем же ключом (409), не от чужого пользователя (у каждого своя область ключей) и не делает запрос «успешным»: временная 5xx не сохраняется как результат, повтор может снова получить 503.
- Область ключа: пользователь + метод + путь чата + ключ; результат хранится ≥ 24 ч (`IDEMPOTENCY_TTL_SECONDS`, меньше 86 400 нельзя).
**Что сломается при увеличении числа API, если сессии находятся в локальной памяти?**
- Вход на реплике A, следующий запрос на B → B сессии не знает → **401**: пользователь «разлогинивается» через раз (у нас запросы одного токена идут по очереди на все реплики, это проверяет тест).
- Logout на A не отзывает токен на B → отозванный токен продолжает работать (нарушение контракта и безопасности).
- Рестарт или rolling update реплики теряет все её сессии.
- Попытка «починить» это sticky-сессиями маскирует проблему: при отказе реплики её пользователи всё равно теряют сессии, а нагрузка распределяется неравномерно. README это запрещает.
- То же касается любой локальной корректностно-важной памяти: кэш истории без инвалидации, локальная идемпотентность (две реплики вставили бы два сообщения), локальные счётчики. У нас в процессе API нет ничего, кроме пула соединений.
## Частые ловушки (на чём преподаватель ловил других) и наш ответ
| Проверка преподавателя | У нас | Тест |
| --- | --- | --- |
| после сбоя ключ остаётся «pending», повтор висит/504/409 | ключ и сообщение в одной транзакции, сбой после первого INSERT → нет ни ключа, ни сообщения, повтор → 201 | `test_failure_after_the_key_insert_rolls_everything_back` |
| одновременные повторы: 19×201 + 1×409 | 24 запроса через ≥ 2 реплики → 24×201, одно тело, одна строка в БД; три раунда | `test_concurrent_same_key_through_different_replicas_creates_exactly_one` |
| посторонний: 404, потом 409 | членство проверяется до ключа: всегда 404 | `test_outsider_and_revoked_session_never_see_the_stored_result` |
| `X-Instance-Id: unknown` / нет на 500 и 413 | hostname контейнера, middleware на всех ответах | `test_every_api_response_carries_instance_and_request_id`, `Api.call` проверяет на каждом ответе |
| LB шлёт запросы неготовой реплике (pg отнят, сеть отключена, пул забит) | активная проверка `/health/ready`, без backup-слотов | `test_lb_stops_routing_…[pg_hba/network]` |
| после общего restart LB видит одну реплику | HAProxy re-resolve Docker DNS каждую 1 с | `test_full_restart_lb_rediscovers_all_replicas` |
| «rolling update» останавливает обе API | +1 новая реплика (healthy и UP), затем graceful stop старой | `bench/rolling_under_load.py`: 0 ошибок |
| миграции в каждой API | только сервис `migrate`, API логов миграции нет | `test_migrations_ran_once_by_the_migrate_service` |
| нет лимитов у LB/migrate/PG/Redis | лимиты у всех 5 сервисов, проверка и config, и реальных контейнеров | `test_compose_policy_…`, `test_running_containers_…` |
| cursor с `[]`, `{}`, числом, `null`, пустой, пробел → 500/200 | 400 на обоих списках | `test_malformed_cursor_values_are_400_on_both_lists` |
| `attachment_ids: []` при повторе → 400 | равно отсутствию поля, тот же 201 | `test_key_order_and_missing_attachment_ids_are_the_same_body` |
| `{"text":"a","TEXT":"b"}`, байт FF, `Bearer é`, UUID в разном регистре, тело при входе | 400 / 400 / 401 / 400 / 400 | `test_api.py`, раздел lab 3 |
## Обнаружение отказа: периоды и максимумы
| Что | Настройка | Максимум |
| --- | --- | --- |
| проверка HAProxy | `inter 1s`, `fastinter 500ms`, `fall 2`, `rise 2`, `timeout check 1s` | — |
| реплика сразу отвечает 503 на ready (drain после SIGTERM) | 1 с до проверки + 0,5 с (`fastinter`), ответ мгновенный | ≈ 1,5 с (измерено ≈ 1,2–1,3 с) |
| ready отвечает медленно: отнят PostgreSQL или Valkey | ответ занимает ≈ 1 с = `timeout check`, поэтому как у замороженной: 1 + 1 + 0,5 + 1 | ≈ 3,5 с (измерено 3,2–4,1 с, включая опрос ≈ 0,1–0,5 с) |
| процесс убит (SIGKILL) / порт закрыт | следующий запрос получает отказ соединения → повтор на другую реплику в HAProxy | новые запросы — сразу; рвутся только запросы «в полёте» (3 запроса за ≤ 0,25 с в серии) |
| процесс заморожен (`docker pause`) | соединения висят до `timeout check` | ≈ 3,5 с до `DOWN`; запросы, попавшие на него раньше, ждут клиентский таймаут |
| DNS: новая/пропавшая реплика | `hold valid 1s` | ≈ 1–2 с |
| graceful stop | readiness 503 сразу, 3 с drain, ≤ 8 с на запросы «в полёте», `stop_grace_period` 15 с | 0 ошибок (серия, тесты) |
## Трек 1 — коротко
Числа и графики — в REPORT.md и [evidence/track1/summary.md](../evidence/track1/summary.md).
- **Методика.** k6 в отдельном контейнере в сети `edge` (как внешний клиент, без проброса портов Windows), open-loop `constant-arrival-rate`, только `201` считаются успехом, `dropped_iterations` — неудовлетворённый спрос. Ступени нагрузки, прогрев 10 с + замер 20 с. На каждую конфигурацию и повтор — **свежий стенд** с пустой БД, образ API собирается из конкретного коммита. CPU закреплены `cpuset`: k6 0–1, LB+Valkey 2, PostgreSQL 3–4, API 5–7. Три повтора.
- **Фиксированный суммарный CPU (2,0):** 1 × 2,0 упирается в одно ядро — один процесс Python исполняет байткод одним потоком (GIL), второе ядро почти простаивает. 2 × 1,0 и 4 × 0,5 используют весь бюджет; дальнейшее дробление не добавляет производительности, только накладные расходы.
- **Рост ресурсов:** реплики помогают, пока не упрёмся в общий ресурс. С БД 0,5 CPU потолок перестаёт расти после 2 реплик — CPU PostgreSQL у предела; добавление CPU БД поднимает потолок, значит, узкое место — БД. LB при этом занят на десятки процентов одного ядра.
- **Где тратится CPU API:** py-spy показал ~57 % времени реплики в драйвере psycopg — накладные расходы на каждый оператор (около 11 обращений к БД на одно сообщение).
- **Улучшение:** вся идемпотентная запись — один вызов SQL-функции `post_message` (миграция 0003) в autocommit. Семантика та же (те же тесты проходят в обоих режимах), A/B на одном образе через `MESSAGE_WRITE_PATH=multi|function`.
+138
View File
@@ -0,0 +1,138 @@
# Запуск на ноутбуке с Windows и сценарий защиты ЛР 3
Команды — в **PowerShell** из папки проекта (где лежит `compose.yaml`). Переменные `$env:...` действуют до закрытия окна.
## 1. Что установить один раз
| Программа | Зачем | Важно |
| --- | --- | --- |
| [Git for Windows](https://git-scm.com/download/win) | clone и Git Bash | путь по умолчанию `C:\Program Files\Git`: оттуда `scripts\*.cmd` запускают `.sh`-хуки |
| [Docker Desktop](https://www.docker.com/products/docker-desktop/) | PostgreSQL, Valkey, HAProxy, API | backend WSL 2, виртуализация в BIOS; для трека 1 в Settings → Resources дать **8 CPU** |
| [Python 3.12+](https://www.python.org/downloads/windows/) | secrets, публичные и свои тесты, скрипты замеров | галочка **Add python.exe to PATH** |
| GNU make (необязательно) | `make check/test/test-failover …` | `winget install ezwinports.make`, затем **новое** окно PowerShell. Рецептам Makefile нужен `sh`: в каждом окне PowerShell один раз `$env:Path += ";C:\Program Files\Git\usr\bin"` (или запускать make из Git Bash). Без make те же команды есть в таблице REPORT.md |
Проверка:
```powershell
git --version; docker version --format '{{.Server.Version}}'; docker compose version; python --version; make --version
```
## 2. Развернуть проект (дома, заранее — нужен интернет)
```powershell
git clone https://gitea.privcloud.tech/wfa/mtusi2026lab3 mtusi2026lab3
cd mtusi2026lab3
git checkout solution/bvt2302-bodanyuk
python scripts\init_secrets.py # случайные пароли в .\secrets (не в Git)
Copy-Item .env.example .env
docker compose --env-file .env config --quiet # конфигурация собирается без ошибок
docker compose --env-file .env up -d --build --wait
```
Первый запуск качает образы (~800 МБ) и собирает API — 3–10 минут. Команда завершается, когда postgres, valkey, обе реплики api и lb в состоянии `healthy`, а `migrate` завершился с кодом 0.
Окружение для своих тестов и скриптов (один раз):
```powershell
python -m venv .venv
.venv\Scripts\python -m pip install --require-hashes --no-deps -r requirements.txt
.venv\Scripts\python -m pip install -r requirements-test.txt -r requirements-dev.txt
$env:PYTHONUTF8 = "1" # публичные тесты читают UTF-8 JSON
```
Проверка в браузере: http://127.0.0.1:8080/health/ready → `{"status":"ok"}`; Swagger — http://127.0.0.1:8080/docs; состояние балансировщика (какие реплики UP/DOWN) — http://127.0.0.1:8404/.
## 3. Перед защитой (в аудитории)
```powershell
cd mtusi2026lab3
$env:PYTHONUTF8 = "1"
$env:Path += ";C:\Program Files\Git\usr\bin" # sh для рецептов Makefile
docker compose up -d --wait # если стенд был остановлен
make check test PYTHON=python # или: python scripts\check_contract.py; python tests\smoke.py
python scripts\lbctl.py status # обе реплики UP
```
## 4. Сценарий защиты из README — по шагам
**Шаг 1. ≥ 2 разных `X-Instance-Id` через LB с одной сессией.**
```powershell
python scripts\lbctl.py replicas # контейнеры api: имя, X-Instance-Id (= hostname), PID, IP
# Один пользователь, один токен, 6 запросов подряд через LB — номера реплик чередуются:
$u = "demo_" + (Get-Random); $p = "Demo-password-123"
Invoke-RestMethod -Method Post http://127.0.0.1:8080/api/v1/users -ContentType 'application/json' -Body (@{username=$u; display_name='Демо'; password=$p} | ConvertTo-Json)
$basic = [Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes("${u}:${p}"))
$t = (Invoke-RestMethod -Method Post http://127.0.0.1:8080/api/v1/auth/sessions -Headers @{Authorization="Basic $basic"}).token
1..6 | ForEach-Object { (Invoke-WebRequest http://127.0.0.1:8080/api/v1/users/me -Headers @{Authorization="Bearer $t"} -UseBasicParsing).Headers['X-Instance-Id'] }
```
Повторный вход не нужен: сессия лежит в общем Valkey. То же автоматически: `.venv\Scripts\python -m pytest student-tests\test_balancing.py -k one_session -v`.
**Шаг 2. Отказ реплики публичным тестом.**
```powershell
make test-failover ACTION_SCRIPT=scripts/stop-one.cmd PYTHON=python
# без make: python tests\resilience.py --mode failover --action-script scripts\stop-one.cmd
```
Тест сам выбирает реплику, обслужившую его первое сообщение, и передаёт её `X-Instance-Id` хуку. Хук останавливает именно этот контейнер и оставляет его остановленным. Вариант «падение процесса» — `$env:STOP_MODE = "kill"` перед командой (SIGKILL), вернуть — `Remove-Item Env:STOP_MODE`.
**Шаг 3. Показать, какой контейнер действительно остановлен, и повторить для второй.**
```powershell
python scripts\lbctl.py replicas # у остановленной: exited, exit_code 0 (stop) или 137 (kill)
python scripts\lbctl.py status # её слот в HAProxy — MAINT/DOWN, вторая UP
docker compose logs --tail 20 api # drain_started -> shutdown_requested -> stopped у остановленной
make restore PYTHON=python # или: & "C:\Program Files\Git\bin\bash.exe" scripts/restore-api.sh; python scripts\lbctl.py wait-up 2
make test-failover ACTION_SCRIPT=scripts/stop-one.cmd PYTHON=python
```
Тест останавливает ту реплику, которая обслужила его первое сообщение, поэтому для второй он может снова выбрать прежнюю. Детерминированно по очереди каждую реплику останавливает серия уровня 4 (~5 мин): `python bench\failover_series.py --output evidence\failover` (stop, kill и pause для каждой реплики, с восстановлением предыдущей).
**Шаг 4. Restart всего стенда с сохранением volumes и прежним токеном.**
```powershell
make test-persistence ACTION_SCRIPT=scripts/restart.cmd PYTHON=python
python scripts\lbctl.py replicas # у всех api свежий started и новый PID; LB снова видит обе реплики
docker compose ps # postgres, valkey, lb — тоже «Up N seconds»
```
**Уровень 4 вживую.**
```powershell
# Неготовая, но живая реплика исключается из LB (pg_hba reject и docker network disconnect), ~1 мин:
.venv\Scripts\python -m pytest student-tests\test_balancing.py -k not_ready -v -s
# Rolling update под нагрузкой (~1,5 мин): новая ревизия на обеих репликах, 0 ошибок
python bench\rolling_under_load.py --output evidence\rolling-update
# Одновременные повторы с одним ключом через разные реплики:
.venv\Scripts\python -m pytest student-tests\test_idempotency.py -k concurrent -v
```
**Трек 1.** Полная серия идёт часы, поэтому на защите — готовые результаты: [evidence/track1/summary.md](../evidence/track1/summary.md) и графики `evidence/track1/*/*/throughput.svg`. Вживую можно показать профиль одной реплики (~2 мин) и одну короткую серию (~6 мин):
```powershell
docker compose stop # основной стенд не должен мешать замеру
python bench\track1\profile.py --label demo --rate 240
python bench\track1\run.py --series demo --label now --rates 150,250,350 --repeats 1 --config replicas=2,MESSAGE_WRITE_PATH=multi --config replicas=2,MESSAGE_WRITE_PATH=function
docker compose up -d --wait # вернуть основной стенд
```
**Все свои тесты** (~8 мин): `.venv\Scripts\python -m pytest student-tests -v`.
## 5. Если что-то пошло не так
| Симптом | Что сделать |
| --- | --- |
| `error during connect … dockerDesktopLinuxEngine` | Docker Desktop не запущен — запустить и подождать минуту |
| `port is already allocated` на 8080/8404 | занято другим стендом (lab2: `docker compose -p mtusi-lab2 stop`) или в `.env` поставить `APP_PUBLISH_PORT=8081`, `LB_STATS_PORT=8405`; тестам — `--base-url http://127.0.0.1:8081` / `$env:BASE_URL` |
| `secrets/... not found` при `up` | не выполнен `python scripts\init_secrets.py` |
| `UnicodeDecodeError` в `tests\*.py` | не задан `$env:PYTHONUTF8 = "1"` |
| `*.cmd`: «не удаётся найти путь» | Git установлен не в `C:\Program Files\Git` — поправить путь в `scripts\*.cmd` |
| `make: command not found` | новое окно PowerShell после установки make или команды без make из REPORT.md |
| make: `'test' is not recognized` | в этом окне не добавлен `C:\Program Files\Git\usr\bin` в `$env:Path` (рецептам нужен `sh`) |
| после `stop-one` в LB одна реплика | так и задумано; вернуть: `make restore PYTHON=python` |
| api `unhealthy` после сна ноутбука | `docker compose restart` и `python scripts\lbctl.py wait-up 2` |
| нужно начать с чистого листа | `docker compose down -v` (удаляет учебные данные) → `docker compose up -d --build --wait` |
Остановить стенд после защиты: `docker compose down` (данные сохранятся) или `docker compose down -v` (данные удаляются).
+266
View File
@@ -0,0 +1,266 @@
# Clean clone: secrets, config, build, migrations from zero, start, versions
# 2026-10-19T15:10:17Z, commit 2708732 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
$ git status --short --branch
## solution/bvt2302-bodanyuk...origin/solution/bvt2302-bodanyuk
[exit 0, 0.1 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe --version
Python 3.14.6
[exit 0, 0.0 s]
$ docker version --format client {{.Client.Version}} / server {{.Server.Version}}
client 29.8.0 / server 29.8.0
[exit 0, 0.1 s]
$ docker compose version
Docker Compose version v5.5.1
[exit 0, 0.1 s]
$ make --version
GNU Make 4.4.1
Built for Windows32
Copyright (C) 1988-2023 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
[exit 0, 0.0 s]
$ python scripts/init_secrets.py
created secrets/postgres_password
created secrets/valkey_password
created secrets/database_url
created secrets/cache_url
[exit 0, 0.1 s]
$ cp .env.example .env
$ docker compose --env-file .env config --quiet
[exit 0, 0.2 s]
$ docker compose --env-file .env up -d --build --wait
#1 [internal] load local bake definitions
#1 reading from stdin 1.36kB 0.1s done
#1 DONE 0.1s
#2 [migrate internal] load build definition from Dockerfile
#2 transferring dockerfile: 2.52kB 0.0s done
#2 DONE 0.0s
#3 [api] resolve image config for docker-image://docker.io/docker/dockerfile:1
#3 DONE 4.0s
#4 [api] docker-image://docker.io/docker/dockerfile:1@sha256:4edf897a3ffa55b89f906fc8cc78afdb3f1834cc9c7083565e611a8a7d5fe99e
#4 resolve docker.io/docker/dockerfile:1@sha256:4edf897a3ffa55b89f906fc8cc78afdb3f1834cc9c7083565e611a8a7d5fe99e 0.1s done
#4 CACHED
#5 [migrate internal] load metadata for docker.io/library/python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
#5 DONE 0.1s
#6 [api internal] load .dockerignore
#6 transferring context: 196B done
#6 DONE 0.0s
#7 [migrate internal] load build context
#7 transferring context: 219.60kB 0.0s done
#7 DONE 0.1s
#8 [api 1/8] FROM docker.io/library/python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
#8 resolve docker.io/library/python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641 0.1s done
#8 DONE 0.1s
#9 [migrate 7/8] COPY migrations/ ./migrations/
#9 CACHED
#10 [migrate 2/8] WORKDIR /app
#10 CACHED
#11 [migrate 3/8] COPY requirements.txt ./
#11 CACHED
#12 [migrate 5/8] RUN groupadd --system --gid 10001 app && useradd --system --uid 10001 --gid app --no-create-home --shell /usr/sbin/nologin app
#12 CACHED
#13 [migrate 4/8] RUN pip install --root-user-action=ignore --require-hashes --no-deps -r requirements.txt
#13 CACHED
#14 [migrate 6/8] COPY src/ ./src/
#14 CACHED
#15 [migrate 8/8] COPY contracts/openapi.json ./contracts/openapi.json
#15 CACHED
#12 [api 5/8] RUN groupadd --system --gid 10001 app && useradd --system --uid 10001 --gid app --no-create-home --shell /usr/sbin/nologin app
#12 CACHED
#14 [api 6/8] COPY src/ ./src/
#14 CACHED
#10 [api 2/8] WORKDIR /app
#10 CACHED
#11 [api 3/8] COPY requirements.txt ./
#11 CACHED
#9 [api 7/8] COPY migrations/ ./migrations/
#9 CACHED
#13 [api 4/8] RUN pip install --root-user-action=ignore --require-hashes --no-deps -r requirements.txt
#13 CACHED
#16 [api] exporting to image
#16 exporting layers 0.0s done
#16 exporting manifest sha256:2e7919b4c60cf9f2135cb5523cf47bc6456f92675261377cb3a312868c357738 0.0s done
#16 exporting config sha256:986458b901447acf8fa6913d7b9548d1fb0b4335f4169158ad6eb2c5e050c1fc 0.0s done
#16 exporting attestation manifest sha256:712b9fac08d59b57dc8931284b52c54ddf116bc5ad39d17e789a98a04499f68d
#16 exporting attestation manifest sha256:712b9fac08d59b57dc8931284b52c54ddf116bc5ad39d17e789a98a04499f68d 0.1s done
#16 exporting manifest list sha256:331feed9c7da600658f524637b0ed8aa919c68e0372b0191d89b3d45e37122e0
#16 exporting manifest list sha256:331feed9c7da600658f524637b0ed8aa919c68e0372b0191d89b3d45e37122e0 0.1s done
#16 naming to docker.io/library/mtusi-messenger:lab3 0.0s done
#16 unpacking to docker.io/library/mtusi-messenger:lab3 0.0s done
#16 DONE 0.4s
#17 [migrate] exporting to image
#17 exporting layers 0.0s done
#17 exporting manifest sha256:06ff1601ca5b94a5b839cf1b1db979c3b1be6f371a81a349c64a49eba4aa718d 0.0s done
#17 exporting config sha256:e415695b978557fe5a23bf1d0a4a95e3e24651ab739b95088dfa124dd115d26a 0.0s done
#17 exporting attestation manifest sha256:3d9641bd01df9b87cdebd4f0b07e8f90504f14b9d3a090c6d8794e121b69c3d7 0.1s done
#17 exporting manifest list sha256:e7940cf7f96b1b5f7c63645c54e1774d32d1408edcfced01dd5b57e93bcb64c4 0.1s done
#17 naming to docker.io/library/mtusi-messenger:lab3 0.0s done
#17 unpacking to docker.io/library/mtusi-messenger:lab3 0.0s done
#17 DONE 0.3s
#18 [migrate] resolving provenance for metadata file
#18 DONE 0.0s
#19 [api] resolving provenance for metadata file
#19 DONE 0.0s
Image mtusi-messenger:lab3 Building
Image mtusi-messenger:lab3 Building
Image mtusi-messenger:lab3 Built
Image mtusi-messenger:lab3 Built
Network mtusi-lab3_edge Creating
Network mtusi-lab3_edge Creating
Volume mtusi-lab3_pgdata Creating
Volume mtusi-lab3_pgdata Creating
Volume mtusi-lab3_valkeydata Creating
Volume mtusi-lab3_valkeydata Creating
Network mtusi-lab3_data Creating
Network mtusi-lab3_data Creating
Volume mtusi-lab3_valkeydata Created
Volume mtusi-lab3_valkeydata Created
Volume mtusi-lab3_pgdata Created
Volume mtusi-lab3_pgdata Created
Network mtusi-lab3_edge Created
Network mtusi-lab3_edge Created
Network mtusi-lab3_data Created
Network mtusi-lab3_data Created
Container mtusi-lab3-postgres-1 Creating
Container mtusi-lab3-valkey-1 Creating
Container mtusi-lab3-valkey-1 Created
Container mtusi-lab3-postgres-1 Created
Container mtusi-lab3-migrate-1 Creating
Container mtusi-lab3-migrate-1 Created
Container mtusi-lab3-api-1 Creating
Container mtusi-lab3-api-2 Creating
Container mtusi-lab3-api-1 Created
Container mtusi-lab3-api-2 Created
Container mtusi-lab3-lb-1 Creating
Container mtusi-lab3-lb-1 Created
Container mtusi-lab3-postgres-1 Starting
Container mtusi-lab3-valkey-1 Starting
Container mtusi-lab3-postgres-1 Started
Container mtusi-lab3-postgres-1 Waiting
Container mtusi-lab3-valkey-1 Started
Container mtusi-lab3-postgres-1 Healthy
Container mtusi-lab3-migrate-1 Starting
Container mtusi-lab3-migrate-1 Started
Container mtusi-lab3-migrate-1 Waiting
Container mtusi-lab3-postgres-1 Waiting
Container mtusi-lab3-valkey-1 Waiting
Container mtusi-lab3-valkey-1 Healthy
Container mtusi-lab3-postgres-1 Healthy
Container mtusi-lab3-migrate-1 Exited
Container mtusi-lab3-api-2 Starting
Container mtusi-lab3-api-2 Started
Container mtusi-lab3-api-1 Starting
Container mtusi-lab3-api-1 Started
Container mtusi-lab3-api-2 Waiting
Container mtusi-lab3-api-1 Waiting
Container mtusi-lab3-api-2 Healthy
Container mtusi-lab3-api-1 Healthy
Container mtusi-lab3-lb-1 Starting
Container mtusi-lab3-lb-1 Started
Container mtusi-lab3-valkey-1 Waiting
Container mtusi-lab3-migrate-1 Waiting
Container mtusi-lab3-api-2 Waiting
Container mtusi-lab3-api-1 Waiting
Container mtusi-lab3-lb-1 Waiting
Container mtusi-lab3-postgres-1 Waiting
Container mtusi-lab3-postgres-1 Healthy
Container mtusi-lab3-valkey-1 Healthy
Container mtusi-lab3-migrate-1 Exited
Container mtusi-lab3-api-2 Healthy
Container mtusi-lab3-api-1 Healthy
Container mtusi-lab3-lb-1 Healthy
[exit 0, 28.2 s]
$ docker compose ps -a --format table {{.Service}} {{.Image}} {{.Status}}
SERVICE IMAGE STATUS
api mtusi-messenger:lab3 Up 11 seconds (healthy)
api mtusi-messenger:lab3 Up 12 seconds (healthy)
lb haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122 Up 6 seconds (healthy)
migrate mtusi-messenger:lab3 Exited (0) 12 seconds ago
postgres postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c Up 20 seconds (healthy)
valkey valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571 Up 19 seconds (healthy)
[exit 0, 0.3 s]
$ docker compose logs --no-log-prefix migrate
migrate: 0001_init: applied
migrate: 0002_idempotency_keys: applied
migrate: 0003_post_message_function: applied
migrate: done, 3 migration file(s) checked
[exit 0, 0.3 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py replicas
mtusi-lab3-api-1 X-Instance-Id=94b38b941a8d running health=healthy pid=23944 started=2026-10-19T15:10:34.257 ip=172.23.0.3,172.24.0.5
mtusi-lab3-api-2 X-Instance-Id=a8d9cffd7a6e running health=healthy pid=23887 started=2026-10-19T15:10:33.970 ip=172.23.0.2,172.24.0.4
[exit 0, 0.7 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py status
api1 172.23.0.3 UP mtusi-lab3-api-1 X-Instance-Id=94b38b941a8d
api2 172.23.0.2 UP mtusi-lab3-api-2 X-Instance-Id=a8d9cffd7a6e
(other slots: MAINT, no container)
[exit 0, 0.9 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe -m venv .venv
[exit 0, 6.9 s]
$ C:\Users\wfa22\Desktop\Высоконагруженные приложения\mtusi2026lab3-clean\.venv\Scripts\python.exe -m pip install -q --require-hashes --no-deps -r requirements.txt
[notice] A new release of pip is available: 26.1.2 -> 26.2.1
[notice] To update, run: C:\Users\wfa22\Desktop\Высоконагруженные приложения\mtusi2026lab3-clean\.venv\Scripts\python.exe -m pip install --upgrade pip
[exit 0, 15.8 s]
$ C:\Users\wfa22\Desktop\Высоконагруженные приложения\mtusi2026lab3-clean\.venv\Scripts\python.exe -m pip install -q -r requirements-test.txt -r requirements-dev.txt
[notice] A new release of pip is available: 26.1.2 -> 26.2.1
[notice] To update, run: C:\Users\wfa22\Desktop\Высоконагруженные приложения\mtusi2026lab3-clean\.venv\Scripts\python.exe -m pip install --upgrade pip
[exit 0, 22.3 s]
$ git check-ignore -v secrets/database_url .env
.gitignore:5:secrets/ secrets/database_url
.gitignore:2:.env .env
$ git status --short # secrets/ и .env не попадают в Git
?? evidence/clean-clone.txt
?? evidence/failover/
?? evidence/level4-runs.txt
?? evidence/load.json
?? evidence/make-check-validate-test.txt
?? evidence/rolling-update/
?? evidence/student-tests.txt
?? evidence/test-failover.txt
?? evidence/test-persistence.txt
[exit 0]
+504
View File
@@ -0,0 +1,504 @@
{
"meta": {
"date": "2026-10-19T15:23:54+00:00",
"commit": "2708732",
"host": "Windows 10, docker 29.8.0",
"workload": "4 closed-loop workers from the host via 127.0.0.1:8080: POST message (Idempotency-Key, uncertain -> same key) + GET last page; client timeout 3 s",
"lb": "HAProxy: check GET /health/ready inter 1s fastinter 500ms fall 2 rise 2, timeout check 1s; retry-on conn-failure + redispatch",
"api": "DRAIN_SECONDS=3, SHUTDOWN_TIMEOUT_SECONDS=8, stop_grace_period 15s"
},
"rounds": [
{
"mode": "stop",
"replica": "mtusi-lab3-api-1",
"instance_id": "907f96037a59",
"stop_command_seconds": 4.03,
"before": {
"seconds": 5.5,
"requests": 1621,
"ok": 1621,
"failed": 0,
"ok_rps": 294.6,
"error_ratio": 0.0,
"statuses": {
"201": 812,
"200": 809
},
"instances": {
"c3f64421973d": 810,
"907f96037a59": 811
},
"p50_ms": 10.88,
"p95_ms": 31.93,
"p99_ms": 60.82
},
"during": {
"seconds": 16.03,
"requests": 3568,
"ok": 3568,
"failed": 0,
"ok_rps": 222.5,
"error_ratio": 0.0,
"statuses": {
"200": 1784,
"201": 1784
},
"instances": {
"c3f64421973d": 3434,
"907f96037a59": 134
},
"p50_ms": 13.82,
"p95_ms": 61.1,
"p99_ms": 71.93
},
"after": {
"seconds": 6.02,
"requests": 1503,
"ok": 1503,
"failed": 0,
"ok_rps": 249.7,
"error_ratio": 0.0,
"statuses": {
"201": 750,
"200": 753
},
"instances": {
"c3f64421973d": 753,
"907f96037a59": 750
},
"p50_ms": 11.19,
"p95_ms": 55.98,
"p99_ms": 63.38
},
"failed_after_stop": 0,
"failed_statuses": [],
"last_failure_after_stop_s": 0.0,
"target_last_served_after_stop_s": 1.283,
"rejoin_up_in_lb_s": 3.25,
"rejoin_serving_s": 3.35,
"client_retries_same_key": 0,
"history": {
"history_messages": 3692,
"confirmed_messages": 3692,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"restarted_state": "running health=healthy"
},
{
"mode": "stop",
"replica": "mtusi-lab3-api-2",
"instance_id": "c3f64421973d",
"stop_command_seconds": 3.79,
"before": {
"seconds": 5.5,
"requests": 1426,
"ok": 1426,
"failed": 0,
"ok_rps": 259.2,
"error_ratio": 0.0,
"statuses": {
"201": 714,
"200": 712
},
"instances": {
"c3f64421973d": 713,
"907f96037a59": 713
},
"p50_ms": 11.18,
"p95_ms": 53.47,
"p99_ms": 64.41
},
"during": {
"seconds": 15.79,
"requests": 3565,
"ok": 3565,
"failed": 0,
"ok_rps": 225.7,
"error_ratio": 0.0,
"statuses": {
"200": 1782,
"201": 1783
},
"instances": {
"c3f64421973d": 222,
"907f96037a59": 3343
},
"p50_ms": 13.63,
"p95_ms": 63.19,
"p99_ms": 71.39
},
"after": {
"seconds": 6.02,
"requests": 1656,
"ok": 1656,
"failed": 0,
"ok_rps": 275.1,
"error_ratio": 0.0,
"statuses": {
"201": 827,
"200": 829
},
"instances": {
"c3f64421973d": 828,
"907f96037a59": 828
},
"p50_ms": 11.19,
"p95_ms": 41.65,
"p99_ms": 60.73
},
"failed_after_stop": 0,
"failed_statuses": [],
"last_failure_after_stop_s": 0.0,
"target_last_served_after_stop_s": 1.235,
"rejoin_up_in_lb_s": 1.35,
"rejoin_serving_s": 3.07,
"client_retries_same_key": 0,
"history": {
"history_messages": 3690,
"confirmed_messages": 3690,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"restarted_state": "running health=healthy"
},
{
"mode": "kill",
"replica": "mtusi-lab3-api-1",
"instance_id": "907f96037a59",
"stop_command_seconds": 0.72,
"before": {
"seconds": 5.5,
"requests": 1491,
"ok": 1491,
"failed": 0,
"ok_rps": 270.9,
"error_ratio": 0.0,
"statuses": {
"201": 745,
"200": 746
},
"instances": {
"907f96037a59": 746,
"c3f64421973d": 745
},
"p50_ms": 11.18,
"p95_ms": 41.28,
"p99_ms": 56.2
},
"during": {
"seconds": 12.72,
"requests": 2734,
"ok": 2731,
"failed": 3,
"ok_rps": 214.7,
"error_ratio": 0.0011,
"statuses": {
"201": 1366,
"200": 1365,
"0": 3
},
"instances": {
"907f96037a59": 22,
"c3f64421973d": 2709
},
"p50_ms": 13.72,
"p95_ms": 59.09,
"p99_ms": 69.77
},
"after": {
"seconds": 6.02,
"requests": 1477,
"ok": 1477,
"failed": 0,
"ok_rps": 245.5,
"error_ratio": 0.0,
"statuses": {
"200": 740,
"201": 737
},
"instances": {
"c3f64421973d": 739,
"907f96037a59": 738
},
"p50_ms": 11.54,
"p95_ms": 54.73,
"p99_ms": 62.68
},
"failed_after_stop": 3,
"failed_statuses": [
0
],
"last_failure_after_stop_s": 0.247,
"target_last_served_after_stop_s": 0.22,
"rejoin_up_in_lb_s": 1.17,
"rejoin_serving_s": 2.8,
"client_retries_same_key": 2,
"history": {
"history_messages": 3200,
"confirmed_messages": 3200,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"restarted_state": "running health=healthy"
},
{
"mode": "kill",
"replica": "mtusi-lab3-api-2",
"instance_id": "c3f64421973d",
"stop_command_seconds": 0.77,
"before": {
"seconds": 5.5,
"requests": 1241,
"ok": 1241,
"failed": 0,
"ok_rps": 225.6,
"error_ratio": 0.0,
"statuses": {
"201": 620,
"200": 621
},
"instances": {
"c3f64421973d": 621,
"907f96037a59": 620
},
"p50_ms": 11.41,
"p95_ms": 57.52,
"p99_ms": 66.91
},
"during": {
"seconds": 12.77,
"requests": 2297,
"ok": 2294,
"failed": 3,
"ok_rps": 179.6,
"error_ratio": 0.00131,
"statuses": {
"200": 1147,
"201": 1147,
"0": 3
},
"instances": {
"907f96037a59": 2276,
"c3f64421973d": 18
},
"p50_ms": 15.06,
"p95_ms": 66.96,
"p99_ms": 77.44
},
"after": {
"seconds": 6.01,
"requests": 1442,
"ok": 1442,
"failed": 0,
"ok_rps": 240.0,
"error_ratio": 0.0,
"statuses": {
"200": 722,
"201": 720
},
"instances": {
"c3f64421973d": 721,
"907f96037a59": 721
},
"p50_ms": 11.64,
"p95_ms": 53.4,
"p99_ms": 65.32
},
"failed_after_stop": 3,
"failed_statuses": [
0
],
"last_failure_after_stop_s": 0.214,
"target_last_served_after_stop_s": 0.174,
"rejoin_up_in_lb_s": 2.32,
"rejoin_serving_s": 3.64,
"client_retries_same_key": 2,
"history": {
"history_messages": 2885,
"confirmed_messages": 2885,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"restarted_state": "running health=healthy"
},
{
"mode": "pause",
"replica": "mtusi-lab3-api-1",
"instance_id": "907f96037a59",
"stop_command_seconds": 0.2,
"before": {
"seconds": 5.5,
"requests": 1493,
"ok": 1493,
"failed": 0,
"ok_rps": 271.3,
"error_ratio": 0.0,
"statuses": {
"200": 747,
"201": 746
},
"instances": {
"907f96037a59": 747,
"c3f64421973d": 746
},
"p50_ms": 11.06,
"p95_ms": 49.85,
"p99_ms": 61.7
},
"during": {
"seconds": 12.2,
"requests": 1653,
"ok": 1649,
"failed": 4,
"ok_rps": 135.2,
"error_ratio": 0.00242,
"statuses": {
"201": 824,
"200": 825,
"0": 4
},
"instances": {
"c3f64421973d": 1637,
"907f96037a59": 12
},
"p50_ms": 15.99,
"p95_ms": 68.99,
"p99_ms": 79.54
},
"after": {
"seconds": 6.03,
"requests": 1474,
"ok": 1474,
"failed": 0,
"ok_rps": 244.6,
"error_ratio": 0.0,
"statuses": {
"201": 736,
"200": 738
},
"instances": {
"907f96037a59": 737,
"c3f64421973d": 737
},
"p50_ms": 12.41,
"p95_ms": 49.6,
"p99_ms": 64.12
},
"failed_after_stop": 4,
"failed_statuses": [
0
],
"last_failure_after_stop_s": 3.204,
"target_last_served_after_stop_s": 0.148,
"rejoin_up_in_lb_s": 2.21,
"rejoin_serving_s": 2.31,
"client_retries_same_key": 3,
"history": {
"history_messages": 2610,
"confirmed_messages": 2610,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"restarted_state": "running health=healthy"
},
{
"mode": "pause",
"replica": "mtusi-lab3-api-2",
"instance_id": "c3f64421973d",
"stop_command_seconds": 0.13,
"before": {
"seconds": 5.5,
"requests": 1395,
"ok": 1395,
"failed": 0,
"ok_rps": 253.6,
"error_ratio": 0.0,
"statuses": {
"200": 697,
"201": 698
},
"instances": {
"c3f64421973d": 698,
"907f96037a59": 697
},
"p50_ms": 12.39,
"p95_ms": 44.63,
"p99_ms": 60.81
},
"during": {
"seconds": 12.13,
"requests": 1787,
"ok": 1783,
"failed": 4,
"ok_rps": 147.0,
"error_ratio": 0.00224,
"statuses": {
"200": 890,
"201": 893,
"0": 4
},
"instances": {
"907f96037a59": 1768,
"c3f64421973d": 15
},
"p50_ms": 15.86,
"p95_ms": 65.74,
"p99_ms": 73.59
},
"after": {
"seconds": 6.02,
"requests": 1522,
"ok": 1522,
"failed": 0,
"ok_rps": 253.0,
"error_ratio": 0.0,
"statuses": {
"200": 762,
"201": 760
},
"instances": {
"c3f64421973d": 761,
"907f96037a59": 761
},
"p50_ms": 12.2,
"p95_ms": 44.57,
"p99_ms": 61.5
},
"failed_after_stop": 4,
"failed_statuses": [
0
],
"last_failure_after_stop_s": 3.12,
"target_last_served_after_stop_s": 0.098,
"rejoin_up_in_lb_s": 1.16,
"rejoin_serving_s": 1.27,
"client_retries_same_key": 1,
"history": {
"history_messages": 2558,
"confirmed_messages": 2558,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"restarted_state": "running health=healthy"
}
]
}
+10
View File
@@ -0,0 +1,10 @@
Дата 2026-10-19T15:23:54+00:00, commit 2708732. 4 closed-loop workers from the host via 127.0.0.1:8080: POST message (Idempotency-Key, uncertain -> same key) + GET last page; client timeout 3 s.
| Режим | Реплика | Запросов во время отказа | Ошибок после stop | ok-RPS до / во время / после | Последняя ошибка, с | Цель ушла из ротации, с | Возврат в LB / обслуживает, с | Потери / дубли |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| stop | mtusi-lab3-api-1 | 3568 | 0 | 294.6 / 222.5 / 249.7 | 0.0 | 1.283 | 3.25 / 3.35 | 0 / 0 |
| stop | mtusi-lab3-api-2 | 3565 | 0 | 259.2 / 225.7 / 275.1 | 0.0 | 1.235 | 1.35 / 3.07 | 0 / 0 |
| kill | mtusi-lab3-api-1 | 2734 | 3 [0] | 270.9 / 214.7 / 245.5 | 0.247 | 0.22 | 1.17 / 2.8 | 0 / 0 |
| kill | mtusi-lab3-api-2 | 2297 | 3 [0] | 225.6 / 179.6 / 240.0 | 0.214 | 0.174 | 2.32 / 3.64 | 0 / 0 |
| pause | mtusi-lab3-api-1 | 1653 | 4 [0] | 271.3 / 135.2 / 244.6 | 3.204 | 0.148 | 2.21 / 2.31 | 0 / 0 |
| pause | mtusi-lab3-api-2 | 1787 | 4 [0] | 253.6 / 147.0 / 253.0 | 3.12 | 0.098 | 1.16 / 1.27 | 0 / 0 |
@@ -0,0 +1,29 @@
second,ok,failed,instances
0,341,0,TARGET:170|c3f64421973d:171
1,195,0,TARGET:98|c3f64421973d:97
2,348,0,TARGET:174|c3f64421973d:174
3,324,0,TARGET:162|c3f64421973d:162
4,173,0,TARGET:86|c3f64421973d:87
5,283,0,TARGET:142|c3f64421973d:141
6,79,3,-:3|TARGET:22|c3f64421973d:57
7,142,0,c3f64421973d:142
8,304,0,c3f64421973d:304
9,226,0,c3f64421973d:226
10,164,0,c3f64421973d:164
11,281,0,c3f64421973d:281
12,261,0,c3f64421973d:261
13,113,0,c3f64421973d:113
14,285,0,c3f64421973d:285
15,260,0,c3f64421973d:260
16,140,0,c3f64421973d:140
17,251,0,c3f64421973d:251
18,290,0,c3f64421973d:290
19,127,0,c3f64421973d:127
20,181,0,c3f64421973d:181
21,315,0,TARGET:88|c3f64421973d:227
22,300,0,TARGET:152|c3f64421973d:148
23,196,0,TARGET:95|c3f64421973d:101
24,194,0,TARGET:97|c3f64421973d:97
25,308,0,TARGET:157|c3f64421973d:151
26,194,0,TARGET:97|c3f64421973d:97
27,126,0,TARGET:61|c3f64421973d:65
1 second ok failed instances
2 0 341 0 TARGET:170|c3f64421973d:171
3 1 195 0 TARGET:98|c3f64421973d:97
4 2 348 0 TARGET:174|c3f64421973d:174
5 3 324 0 TARGET:162|c3f64421973d:162
6 4 173 0 TARGET:86|c3f64421973d:87
7 5 283 0 TARGET:142|c3f64421973d:141
8 6 79 3 -:3|TARGET:22|c3f64421973d:57
9 7 142 0 c3f64421973d:142
10 8 304 0 c3f64421973d:304
11 9 226 0 c3f64421973d:226
12 10 164 0 c3f64421973d:164
13 11 281 0 c3f64421973d:281
14 12 261 0 c3f64421973d:261
15 13 113 0 c3f64421973d:113
16 14 285 0 c3f64421973d:285
17 15 260 0 c3f64421973d:260
18 16 140 0 c3f64421973d:140
19 17 251 0 c3f64421973d:251
20 18 290 0 c3f64421973d:290
21 19 127 0 c3f64421973d:127
22 20 181 0 c3f64421973d:181
23 21 315 0 TARGET:88|c3f64421973d:227
24 22 300 0 TARGET:152|c3f64421973d:148
25 23 196 0 TARGET:95|c3f64421973d:101
26 24 194 0 TARGET:97|c3f64421973d:97
27 25 308 0 TARGET:157|c3f64421973d:151
28 26 194 0 TARGET:97|c3f64421973d:97
29 27 126 0 TARGET:61|c3f64421973d:65
@@ -0,0 +1,30 @@
second,ok,failed,instances
0,299,0,907f96037a59:151|TARGET:148
1,191,0,907f96037a59:92|TARGET:99
2,274,0,907f96037a59:139|TARGET:135
3,335,0,907f96037a59:169|TARGET:166
4,191,0,907f96037a59:92|TARGET:99
5,129,0,907f96037a59:66|TARGET:63
6,61,3,-:3|907f96037a59:43|TARGET:18
7,98,0,907f96037a59:98
8,94,0,907f96037a59:94
9,225,0,907f96037a59:225
10,288,0,907f96037a59:288
11,174,0,907f96037a59:174
12,212,0,907f96037a59:212
13,274,0,907f96037a59:274
14,199,0,907f96037a59:199
15,118,0,907f96037a59:118
16,240,0,907f96037a59:240
17,220,0,907f96037a59:220
18,119,0,907f96037a59:119
19,202,0,907f96037a59:202
20,198,0,907f96037a59:198
21,114,0,907f96037a59:114
22,288,0,907f96037a59:166|TARGET:122
23,323,0,907f96037a59:163|TARGET:160
24,176,0,907f96037a59:85|TARGET:91
25,179,0,907f96037a59:91|TARGET:88
26,284,0,907f96037a59:143|TARGET:141
27,179,0,907f96037a59:87|TARGET:92
28,87,0,907f96037a59:45|TARGET:42
1 second ok failed instances
2 0 299 0 907f96037a59:151|TARGET:148
3 1 191 0 907f96037a59:92|TARGET:99
4 2 274 0 907f96037a59:139|TARGET:135
5 3 335 0 907f96037a59:169|TARGET:166
6 4 191 0 907f96037a59:92|TARGET:99
7 5 129 0 907f96037a59:66|TARGET:63
8 6 61 3 -:3|907f96037a59:43|TARGET:18
9 7 98 0 907f96037a59:98
10 8 94 0 907f96037a59:94
11 9 225 0 907f96037a59:225
12 10 288 0 907f96037a59:288
13 11 174 0 907f96037a59:174
14 12 212 0 907f96037a59:212
15 13 274 0 907f96037a59:274
16 14 199 0 907f96037a59:199
17 15 118 0 907f96037a59:118
18 16 240 0 907f96037a59:240
19 17 220 0 907f96037a59:220
20 18 119 0 907f96037a59:119
21 19 202 0 907f96037a59:202
22 20 198 0 907f96037a59:198
23 21 114 0 907f96037a59:114
24 22 288 0 907f96037a59:166|TARGET:122
25 23 323 0 907f96037a59:163|TARGET:160
26 24 176 0 907f96037a59:85|TARGET:91
27 25 179 0 907f96037a59:91|TARGET:88
28 26 284 0 907f96037a59:143|TARGET:141
29 27 179 0 907f96037a59:87|TARGET:92
30 28 87 0 907f96037a59:45|TARGET:42
@@ -0,0 +1,26 @@
second,ok,failed,instances
0,326,0,TARGET:165|c3f64421973d:161
1,175,0,TARGET:84|c3f64421973d:91
2,329,0,TARGET:166|c3f64421973d:163
3,352,0,TARGET:178|c3f64421973d:174
4,179,0,TARGET:86|c3f64421973d:93
5,294,0,TARGET:149|c3f64421973d:145
6,30,0,TARGET:13|c3f64421973d:17
9,123,4,-:4|c3f64421973d:123
10,148,0,c3f64421973d:148
11,254,0,c3f64421973d:254
12,246,0,c3f64421973d:246
13,118,0,c3f64421973d:118
14,253,0,c3f64421973d:253
15,223,0,c3f64421973d:223
16,67,0,c3f64421973d:67
17,139,0,c3f64421973d:139
18,189,0,TARGET:15|c3f64421973d:174
19,227,0,TARGET:115|c3f64421973d:112
20,160,0,TARGET:77|c3f64421973d:83
21,258,0,TARGET:131|c3f64421973d:127
22,326,0,TARGET:163|c3f64421973d:163
23,167,0,TARGET:82|c3f64421973d:85
24,254,0,TARGET:128|c3f64421973d:126
25,301,0,TARGET:151|c3f64421973d:150
26,84,0,TARGET:42|c3f64421973d:42
1 second ok failed instances
2 0 326 0 TARGET:165|c3f64421973d:161
3 1 175 0 TARGET:84|c3f64421973d:91
4 2 329 0 TARGET:166|c3f64421973d:163
5 3 352 0 TARGET:178|c3f64421973d:174
6 4 179 0 TARGET:86|c3f64421973d:93
7 5 294 0 TARGET:149|c3f64421973d:145
8 6 30 0 TARGET:13|c3f64421973d:17
9 9 123 4 -:4|c3f64421973d:123
10 10 148 0 c3f64421973d:148
11 11 254 0 c3f64421973d:254
12 12 246 0 c3f64421973d:246
13 13 118 0 c3f64421973d:118
14 14 253 0 c3f64421973d:253
15 15 223 0 c3f64421973d:223
16 16 67 0 c3f64421973d:67
17 17 139 0 c3f64421973d:139
18 18 189 0 TARGET:15|c3f64421973d:174
19 19 227 0 TARGET:115|c3f64421973d:112
20 20 160 0 TARGET:77|c3f64421973d:83
21 21 258 0 TARGET:131|c3f64421973d:127
22 22 326 0 TARGET:163|c3f64421973d:163
23 23 167 0 TARGET:82|c3f64421973d:85
24 24 254 0 TARGET:128|c3f64421973d:126
25 25 301 0 TARGET:151|c3f64421973d:150
26 26 84 0 TARGET:42|c3f64421973d:42
@@ -0,0 +1,25 @@
second,ok,failed,instances
0,304,0,907f96037a59:152|TARGET:152
1,248,0,907f96037a59:125|TARGET:123
2,192,0,907f96037a59:95|TARGET:97
3,334,0,907f96037a59:167|TARGET:167
4,280,0,907f96037a59:141|TARGET:139
5,197,0,907f96037a59:97|TARGET:100
6,34,0,907f96037a59:19|TARGET:15
9,181,4,-:4|907f96037a59:181
10,173,0,907f96037a59:173
11,169,0,907f96037a59:169
12,268,0,907f96037a59:268
13,191,0,907f96037a59:191
14,133,0,907f96037a59:133
15,250,0,907f96037a59:250
16,241,0,907f96037a59:241
17,113,0,907f96037a59:113
18,166,0,907f96037a59:145|TARGET:21
19,308,0,907f96037a59:154|TARGET:154
20,167,0,907f96037a59:85|TARGET:82
21,260,0,907f96037a59:129|TARGET:131
22,303,0,907f96037a59:151|TARGET:152
23,211,0,907f96037a59:107|TARGET:104
24,261,0,907f96037a59:129|TARGET:132
25,130,0,907f96037a59:65|TARGET:65
1 second ok failed instances
2 0 304 0 907f96037a59:152|TARGET:152
3 1 248 0 907f96037a59:125|TARGET:123
4 2 192 0 907f96037a59:95|TARGET:97
5 3 334 0 907f96037a59:167|TARGET:167
6 4 280 0 907f96037a59:141|TARGET:139
7 5 197 0 907f96037a59:97|TARGET:100
8 6 34 0 907f96037a59:19|TARGET:15
9 9 181 4 -:4|907f96037a59:181
10 10 173 0 907f96037a59:173
11 11 169 0 907f96037a59:169
12 12 268 0 907f96037a59:268
13 13 191 0 907f96037a59:191
14 14 133 0 907f96037a59:133
15 15 250 0 907f96037a59:250
16 16 241 0 907f96037a59:241
17 17 113 0 907f96037a59:113
18 18 166 0 907f96037a59:145|TARGET:21
19 19 308 0 907f96037a59:154|TARGET:154
20 20 167 0 907f96037a59:85|TARGET:82
21 21 260 0 907f96037a59:129|TARGET:131
22 22 303 0 907f96037a59:151|TARGET:152
23 23 211 0 907f96037a59:107|TARGET:104
24 24 261 0 907f96037a59:129|TARGET:132
25 25 130 0 907f96037a59:65|TARGET:65
@@ -0,0 +1,33 @@
second,ok,failed,instances
0,171,0,TARGET:85|c3f64421973d:86
1,368,0,TARGET:185|c3f64421973d:183
2,272,0,TARGET:137|c3f64421973d:135
3,187,0,TARGET:92|c3f64421973d:95
4,370,0,TARGET:185|c3f64421973d:185
5,330,0,TARGET:166|c3f64421973d:164
6,180,0,TARGET:89|c3f64421973d:91
7,294,0,TARGET:46|c3f64421973d:248
8,227,0,c3f64421973d:227
9,109,0,c3f64421973d:109
10,218,0,c3f64421973d:218
11,277,0,c3f64421973d:277
12,164,0,c3f64421973d:164
13,243,0,c3f64421973d:243
14,298,0,c3f64421973d:298
15,181,0,c3f64421973d:181
16,223,0,c3f64421973d:223
17,292,0,c3f64421973d:292
18,156,0,c3f64421973d:156
19,197,0,c3f64421973d:197
20,277,0,c3f64421973d:277
21,224,0,c3f64421973d:224
22,108,0,c3f64421973d:108
23,207,0,c3f64421973d:207
24,229,0,TARGET:15|c3f64421973d:214
25,209,0,TARGET:105|c3f64421973d:104
26,343,0,TARGET:171|c3f64421973d:172
27,217,0,TARGET:107|c3f64421973d:110
28,168,0,TARGET:86|c3f64421973d:82
29,286,0,TARGET:142|c3f64421973d:144
30,273,0,TARGET:136|c3f64421973d:137
31,86,0,TARGET:44|c3f64421973d:42
1 second ok failed instances
2 0 171 0 TARGET:85|c3f64421973d:86
3 1 368 0 TARGET:185|c3f64421973d:183
4 2 272 0 TARGET:137|c3f64421973d:135
5 3 187 0 TARGET:92|c3f64421973d:95
6 4 370 0 TARGET:185|c3f64421973d:185
7 5 330 0 TARGET:166|c3f64421973d:164
8 6 180 0 TARGET:89|c3f64421973d:91
9 7 294 0 TARGET:46|c3f64421973d:248
10 8 227 0 c3f64421973d:227
11 9 109 0 c3f64421973d:109
12 10 218 0 c3f64421973d:218
13 11 277 0 c3f64421973d:277
14 12 164 0 c3f64421973d:164
15 13 243 0 c3f64421973d:243
16 14 298 0 c3f64421973d:298
17 15 181 0 c3f64421973d:181
18 16 223 0 c3f64421973d:223
19 17 292 0 c3f64421973d:292
20 18 156 0 c3f64421973d:156
21 19 197 0 c3f64421973d:197
22 20 277 0 c3f64421973d:277
23 21 224 0 c3f64421973d:224
24 22 108 0 c3f64421973d:108
25 23 207 0 c3f64421973d:207
26 24 229 0 TARGET:15|c3f64421973d:214
27 25 209 0 TARGET:105|c3f64421973d:104
28 26 343 0 TARGET:171|c3f64421973d:172
29 27 217 0 TARGET:107|c3f64421973d:110
30 28 168 0 TARGET:86|c3f64421973d:82
31 29 286 0 TARGET:142|c3f64421973d:144
32 30 273 0 TARGET:136|c3f64421973d:137
33 31 86 0 TARGET:44|c3f64421973d:42
@@ -0,0 +1,32 @@
second,ok,failed,instances
0,360,0,907f96037a59:180|TARGET:180
1,187,0,907f96037a59:93|TARGET:94
2,271,0,907f96037a59:136|TARGET:135
3,368,0,907f96037a59:184|TARGET:184
4,240,0,907f96037a59:120|TARGET:120
5,182,0,907f96037a59:91|TARGET:91
6,358,0,907f96037a59:179|TARGET:179
7,199,0,907f96037a59:156|TARGET:43
8,158,0,907f96037a59:158
9,249,0,907f96037a59:249
10,245,0,907f96037a59:245
11,129,0,907f96037a59:129
12,291,0,907f96037a59:291
13,291,0,907f96037a59:291
14,127,0,907f96037a59:127
15,224,0,907f96037a59:224
16,295,0,907f96037a59:295
17,139,0,907f96037a59:139
18,238,0,907f96037a59:238
19,299,0,907f96037a59:299
20,165,0,907f96037a59:165
21,207,0,907f96037a59:207
22,193,0,907f96037a59:193
23,170,0,907f96037a59:170
24,189,0,907f96037a59:152|TARGET:37
25,314,0,907f96037a59:157|TARGET:157
26,281,0,907f96037a59:139|TARGET:142
27,185,0,907f96037a59:94|TARGET:91
28,318,0,907f96037a59:159|TARGET:159
29,338,0,907f96037a59:167|TARGET:171
30,170,0,907f96037a59:87|TARGET:83
1 second ok failed instances
2 0 360 0 907f96037a59:180|TARGET:180
3 1 187 0 907f96037a59:93|TARGET:94
4 2 271 0 907f96037a59:136|TARGET:135
5 3 368 0 907f96037a59:184|TARGET:184
6 4 240 0 907f96037a59:120|TARGET:120
7 5 182 0 907f96037a59:91|TARGET:91
8 6 358 0 907f96037a59:179|TARGET:179
9 7 199 0 907f96037a59:156|TARGET:43
10 8 158 0 907f96037a59:158
11 9 249 0 907f96037a59:249
12 10 245 0 907f96037a59:245
13 11 129 0 907f96037a59:129
14 12 291 0 907f96037a59:291
15 13 291 0 907f96037a59:291
16 14 127 0 907f96037a59:127
17 15 224 0 907f96037a59:224
18 16 295 0 907f96037a59:295
19 17 139 0 907f96037a59:139
20 18 238 0 907f96037a59:238
21 19 299 0 907f96037a59:299
22 20 165 0 907f96037a59:165
23 21 207 0 907f96037a59:207
24 22 193 0 907f96037a59:193
25 23 170 0 907f96037a59:170
26 24 189 0 907f96037a59:152|TARGET:37
27 25 314 0 907f96037a59:157|TARGET:157
28 26 281 0 907f96037a59:139|TARGET:142
29 27 185 0 907f96037a59:94|TARGET:91
30 28 318 0 907f96037a59:159|TARGET:159
31 29 338 0 907f96037a59:167|TARGET:171
32 30 170 0 907f96037a59:87|TARGET:83
+84
View File
@@ -0,0 +1,84 @@
# Level 4: failover series, rolling update under load, public load.py
# 2026-10-19T15:20:55Z, commit 2708732 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe bench/failover_series.py --output evidence/failover
== stop mtusi-lab3-api-1 ...
{"failed_after_stop": 0, "last_failure_after_stop_s": 0.0, "target_last_served_after_stop_s": 1.283, "rejoin_serving_s": 3.35}
== stop mtusi-lab3-api-2 ...
{"failed_after_stop": 0, "last_failure_after_stop_s": 0.0, "target_last_served_after_stop_s": 1.235, "rejoin_serving_s": 3.07}
== kill mtusi-lab3-api-1 ...
{"failed_after_stop": 3, "last_failure_after_stop_s": 0.247, "target_last_served_after_stop_s": 0.22, "rejoin_serving_s": 2.8}
== kill mtusi-lab3-api-2 ...
{"failed_after_stop": 3, "last_failure_after_stop_s": 0.214, "target_last_served_after_stop_s": 0.174, "rejoin_serving_s": 3.64}
== pause mtusi-lab3-api-1 ...
{"failed_after_stop": 4, "last_failure_after_stop_s": 3.204, "target_last_served_after_stop_s": 0.148, "rejoin_serving_s": 2.31}
== pause mtusi-lab3-api-2 ...
{"failed_after_stop": 4, "last_failure_after_stop_s": 3.12, "target_last_served_after_stop_s": 0.098, "rejoin_serving_s": 1.27}
| Режим | Реплика | Запросов во время отказа | Ошибок после stop | ok-RPS до / во время / после | Последняя ошибка, с | Цель ушла из ротации, с | Возврат в LB / обслуживает, с | Потери / дубли |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| stop | mtusi-lab3-api-1 | 3568 | 0 | 294.6 / 222.5 / 249.7 | 0.0 | 1.283 | 3.25 / 3.35 | 0 / 0 |
| stop | mtusi-lab3-api-2 | 3565 | 0 | 259.2 / 225.7 / 275.1 | 0.0 | 1.235 | 1.35 / 3.07 | 0 / 0 |
| kill | mtusi-lab3-api-1 | 2734 | 3 [0] | 270.9 / 214.7 / 245.5 | 0.247 | 0.22 | 1.17 / 2.8 | 0 / 0 |
| kill | mtusi-lab3-api-2 | 2297 | 3 [0] | 225.6 / 179.6 / 240.0 | 0.214 | 0.174 | 2.32 / 3.64 | 0 / 0 |
| pause | mtusi-lab3-api-1 | 1653 | 4 [0] | 271.3 / 135.2 / 244.6 | 3.204 | 0.148 | 2.21 / 2.31 | 0 / 0 |
| pause | mtusi-lab3-api-2 | 1787 | 4 [0] | 253.6 / 147.0 / 253.0 | 3.12 | 0.098 | 1.16 / 1.27 | 0 / 0 |
[exit 0, 179.0 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe bench/rolling_under_load.py --output evidence/rolling-update
{
"script_exit_code": 0,
"update_seconds": 57.1,
"before": {
"mtusi-lab3-api-1": "907f96037a59 revision=dev",
"mtusi-lab3-api-2": "c3f64421973d revision=dev"
},
"after": {
"mtusi-lab3-api-3": "97c3063d8125 revision=2708732-20261019T152401Z",
"mtusi-lab3-api-4": "7a0d24d4d31c revision=2708732-20261019T152401Z"
},
"history": {
"history_messages": 7119,
"confirmed_messages": 7119,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
}
}
{"seconds": 57.09, "requests": 11848, "ok": 11848, "failed": 0, "ok_rps": 207.5, "error_ratio": 0.0, "statuses": {"201": 5924, "200": 5924}, "instances": {"907f96037a59": 3080, "c3f64421973d": 4767, "97c3063d8125": 2891, "7a0d24d4d31c": 1110}, "p50_ms": 14.32, "p95_ms": 53.01, "p99_ms": 68.28}
[exit 0, 74.9 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py wait-up 2
2 API replica(s) UP in HAProxy after 0.7 s: ['7a0d24d4d31c', '97c3063d8125']
[exit 0, 1.7 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe tests/load.py --duration 30 --concurrency 4 --output evidence/load.json
{
"lab": 3,
"workload": "closed-loop POST messages; no automatic retries",
"requested_duration_seconds": 30.0,
"elapsed_seconds": 30.093,
"concurrency": 4,
"successful_requests": 5247,
"failed_requests": 0,
"error_ratio": 0.0,
"successful_rps": 174.357,
"successful_latency_ms": {
"p50": 18.617,
"p95": 50.067,
"p99": 65.226
},
"note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead."
}
[exit 0, 30.8 s]
$ docker compose ps -a --format table {{.Service}} {{.Status}}
SERVICE STATUS
api Up About a minute (healthy)
api Up About a minute (healthy)
lb Up 9 minutes (healthy)
migrate Exited (0) 4 minutes ago
postgres Up 5 minutes (healthy)
valkey Up 6 minutes (healthy)
[exit 0, 0.3 s]
+17
View File
@@ -0,0 +1,17 @@
{
"lab": 3,
"workload": "closed-loop POST messages; no automatic retries",
"requested_duration_seconds": 30.0,
"elapsed_seconds": 30.093,
"concurrency": 4,
"successful_requests": 5247,
"failed_requests": 0,
"error_ratio": 0.0,
"successful_rps": 174.357,
"successful_latency_ms": {
"p50": 18.617,
"p95": 50.067,
"p99": 65.226
},
"note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead."
}
+17
View File
@@ -0,0 +1,17 @@
# make check, make validate (venv), make test
# 2026-10-19T15:11:33Z, commit 2708732 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
$ make check
python scripts/check_contract.py
PASS: offline skeleton structure, local refs and Python syntax; application NOT tested.
[exit 0, 0.6 s]
$ make validate PYTHON=.venv/Scripts/python.exe
.venv/Scripts/python.exe scripts/validate_openapi.py
PASS: full OpenAPI 3.1 and JSON Schema validation; application NOT tested.
[exit 0, 1.9 s]
$ make test
python tests/smoke.py --base-url "http://localhost:8080"
PASS: lab3 public API scenario; infrastructure and grading evidence remain separate.
[exit 0, 1.2 s]
+108
View File
@@ -0,0 +1,108 @@
$ sh scripts/rolling-update.sh (exit 0)
rolling-update: 2 replica(s) to replace, new revision 2708732-20261019T152401Z
#1 [internal] load local bake definitions
#1 reading from stdin 1.38kB 0.1s done
#1 DONE 0.1s
#2 [internal] load build definition from Dockerfile
#2 transferring dockerfile: 2.52kB 0.0s done
#2 DONE 0.0s
#3 resolve image config for docker-image://docker.io/docker/dockerfile:1
#3 DONE 4.9s
#4 docker-image://docker.io/docker/dockerfile:1@sha256:4edf897a3ffa55b89f906fc8cc78afdb3f1834cc9c7083565e611a8a7d5fe99e
#4 resolve docker.io/docker/dockerfile:1@sha256:4edf897a3ffa55b89f906fc8cc78afdb3f1834cc9c7083565e611a8a7d5fe99e 0.1s done
#4 CACHED
#5 [internal] load metadata for docker.io/library/python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
#5 DONE 3.3s
#6 [internal] load .dockerignore
#6 transferring context: 196B done
#6 DONE 0.0s
#7 [internal] load build context
#7 transferring context: 1.11kB 0.0s done
#7 DONE 0.1s
#8 [1/8] FROM docker.io/library/python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641
#8 resolve docker.io/library/python:3.13.16-slim-bookworm@sha256:a1165e272e578941b84abc79e4ab38a0305cd12803a5c4247979ac7655f4d641 0.1s done
#8 DONE 0.1s
#9 [6/8] COPY src/ ./src/
#9 CACHED
#10 [7/8] COPY migrations/ ./migrations/
#10 CACHED
#11 [2/8] WORKDIR /app
#11 CACHED
#12 [4/8] RUN pip install --root-user-action=ignore --require-hashes --no-deps -r requirements.txt
#12 CACHED
#13 [3/8] COPY requirements.txt ./
#13 CACHED
#14 [5/8] RUN groupadd --system --gid 10001 app && useradd --system --uid 10001 --gid app --no-create-home --shell /usr/sbin/nologin app
#14 CACHED
#15 [8/8] COPY contracts/openapi.json ./contracts/openapi.json
#15 CACHED
#16 exporting to image
#16 exporting layers 0.0s done
#16 exporting manifest sha256:42a5367c0c23b0291d73d5a6e1b863c547c88f1acc80d90c707850b79e894d33
#16 exporting manifest sha256:42a5367c0c23b0291d73d5a6e1b863c547c88f1acc80d90c707850b79e894d33 0.0s done
#16 exporting config sha256:45ff09fc4f0a55a942e45f56fcdf19071c8c76fd5a4a0243912407c23439bbe7 0.1s done
#16 exporting attestation manifest sha256:92b52b37c33ce3d4c52d0aef43d53bae0fe620968c5277e0a0905e6f1443822c 0.1s done
#16 exporting manifest list sha256:ccb3adf721f08abed4b0678ad0c49e587a0871f560ad4dfc686957846825cf9a
#16 exporting manifest list sha256:ccb3adf721f08abed4b0678ad0c49e587a0871f560ad4dfc686957846825cf9a 0.1s done
#16 naming to docker.io/library/mtusi-messenger:lab3 0.0s done
#16 unpacking to docker.io/library/mtusi-messenger:lab3
#16 unpacking to docker.io/library/mtusi-messenger:lab3 0.1s done
#16 DONE 0.6s
#17 resolving provenance for metadata file
#17 DONE 0.0s
migrate: 0001_init: already applied, skipped
migrate: 0002_idempotency_keys: already applied, skipped
migrate: 0003_post_message_function: already applied, skipped
migrate: done, 3 migration file(s) checked
3 API replica(s) UP in HAProxy after 1.0 s: ['907f96037a59', '97c3063d8125', 'c3f64421973d']
2 API replica(s) UP in HAProxy after 1.0 s: ['97c3063d8125', 'c3f64421973d']
rolling-update: replaced mtusi-lab3-api-1
3 API replica(s) UP in HAProxy after 1.4 s: ['7a0d24d4d31c', '97c3063d8125', 'c3f64421973d']
2 API replica(s) UP in HAProxy after 0.8 s: ['7a0d24d4d31c', '97c3063d8125']
rolling-update: replaced mtusi-lab3-api-2
/mtusi-lab3-api-3 revision=2708732-20261019T152401Z running
/mtusi-lab3-api-4 revision=2708732-20261019T152401Z running
Image mtusi-messenger:lab3 Building
Image mtusi-messenger:lab3 Built
Container mtusi-lab3-migrate-run-5a1b1047bd22 Creating
Container mtusi-lab3-migrate-run-5a1b1047bd22 Created
Container mtusi-lab3-api-1 Running
Container mtusi-lab3-api-2 Running
Container mtusi-lab3-api-3 Creating
Container mtusi-lab3-api-3 Created
Container mtusi-lab3-api-3 Starting
Container mtusi-lab3-api-3 Started
Container mtusi-lab3-api-3 Waiting
Container mtusi-lab3-api-2 Waiting
Container mtusi-lab3-api-1 Waiting
Container mtusi-lab3-api-3 Healthy
Container mtusi-lab3-api-2 Healthy
Container mtusi-lab3-api-1 Healthy
Container mtusi-lab3-api-2 Running
Container mtusi-lab3-api-3 Running
Container mtusi-lab3-api-4 Creating
Container mtusi-lab3-api-4 Created
Container mtusi-lab3-api-4 Starting
Container mtusi-lab3-api-4 Started
Container mtusi-lab3-api-4 Waiting
Container mtusi-lab3-api-3 Waiting
Container mtusi-lab3-api-2 Waiting
Container mtusi-lab3-api-4 Healthy
Container mtusi-lab3-api-3 Healthy
Container mtusi-lab3-api-2 Healthy
+65
View File
@@ -0,0 +1,65 @@
{
"script_exit_code": 0,
"update_seconds": 57.1,
"before": {
"mtusi-lab3-api-1": "907f96037a59 revision=dev",
"mtusi-lab3-api-2": "c3f64421973d revision=dev"
},
"after": {
"mtusi-lab3-api-3": "97c3063d8125 revision=2708732-20261019T152401Z",
"mtusi-lab3-api-4": "7a0d24d4d31c revision=2708732-20261019T152401Z"
},
"during_update": {
"seconds": 57.09,
"requests": 11848,
"ok": 11848,
"failed": 0,
"ok_rps": 207.5,
"error_ratio": 0.0,
"statuses": {
"201": 5924,
"200": 5924
},
"instances": {
"907f96037a59": 3080,
"c3f64421973d": 4767,
"97c3063d8125": 2891,
"7a0d24d4d31c": 1110
},
"p50_ms": 14.32,
"p95_ms": 53.01,
"p99_ms": 68.28
},
"whole_run": {
"seconds": 67.79,
"requests": 14238,
"ok": 14238,
"failed": 0,
"ok_rps": 210.0,
"error_ratio": 0.0,
"statuses": {
"201": 7119,
"200": 7119
},
"instances": {
"907f96037a59": 3754,
"c3f64421973d": 5445,
"97c3063d8125": 3410,
"7a0d24d4d31c": 1629
},
"p50_ms": 13.97,
"p95_ms": 52.62,
"p99_ms": 67.65
},
"client_retries_same_key": 0,
"history": {
"history_messages": 7119,
"confirmed_messages": 7119,
"lost_confirmed": 0,
"altered_confirmed": 0,
"duplicate_ids": 0,
"duplicate_texts": 0,
"unconfirmed_but_stored": 0
},
"workload": "4 closed-loop workers: POST (Idempotency-Key) + GET page via 127.0.0.1:8080"
}
+69
View File
@@ -0,0 +1,69 @@
second,ok,failed,instances
0,324,0,907f96037a59:162|c3f64421973d:162
1,218,0,907f96037a59:108|c3f64421973d:110
2,188,0,907f96037a59:95|c3f64421973d:93
3,313,0,907f96037a59:156|c3f64421973d:157
4,309,0,907f96037a59:153|c3f64421973d:156
5,132,0,907f96037a59:67|c3f64421973d:65
6,224,0,907f96037a59:113|c3f64421973d:111
7,250,0,907f96037a59:125|c3f64421973d:125
8,176,0,907f96037a59:86|c3f64421973d:90
9,232,0,907f96037a59:118|c3f64421973d:114
10,337,0,907f96037a59:168|c3f64421973d:169
11,181,0,907f96037a59:89|c3f64421973d:92
12,227,0,907f96037a59:115|c3f64421973d:112
13,290,0,907f96037a59:145|c3f64421973d:145
14,277,0,907f96037a59:138|c3f64421973d:139
15,204,0,907f96037a59:103|c3f64421973d:101
16,302,0,907f96037a59:151|c3f64421973d:151
17,203,0,907f96037a59:100|c3f64421973d:103
18,108,0,907f96037a59:54|c3f64421973d:54
19,216,0,907f96037a59:110|c3f64421973d:106
20,252,0,907f96037a59:125|c3f64421973d:127
21,145,0,907f96037a59:72|c3f64421973d:73
22,214,0,907f96037a59:108|c3f64421973d:106
23,231,0,907f96037a59:115|c3f64421973d:116
24,187,0,907f96037a59:92|c3f64421973d:95
25,138,0,907f96037a59:69|c3f64421973d:69
26,237,0,907f96037a59:120|c3f64421973d:117
27,249,0,907f96037a59:125|c3f64421973d:124
28,188,0,907f96037a59:77|97c3063d8125:34|c3f64421973d:77
29,334,0,907f96037a59:111|97c3063d8125:111|c3f64421973d:112
30,289,0,907f96037a59:96|97c3063d8125:97|c3f64421973d:96
31,139,0,907f96037a59:45|97c3063d8125:47|c3f64421973d:47
32,232,0,907f96037a59:79|97c3063d8125:77|c3f64421973d:76
33,240,0,907f96037a59:80|97c3063d8125:79|c3f64421973d:81
34,170,0,907f96037a59:55|97c3063d8125:58|c3f64421973d:57
35,260,0,907f96037a59:29|97c3063d8125:116|c3f64421973d:115
36,334,0,97c3063d8125:166|c3f64421973d:168
37,203,0,97c3063d8125:103|c3f64421973d:100
38,176,0,97c3063d8125:87|c3f64421973d:89
39,232,0,97c3063d8125:116|c3f64421973d:116
40,182,0,97c3063d8125:92|c3f64421973d:90
41,88,0,97c3063d8125:44|c3f64421973d:44
42,202,0,97c3063d8125:100|c3f64421973d:102
43,225,0,97c3063d8125:112|c3f64421973d:113
44,171,0,97c3063d8125:87|c3f64421973d:84
45,197,0,97c3063d8125:97|c3f64421973d:100
46,342,0,7a0d24d4d31c:9|97c3063d8125:167|c3f64421973d:166
47,195,0,7a0d24d4d31c:65|97c3063d8125:65|c3f64421973d:65
48,134,0,7a0d24d4d31c:45|97c3063d8125:45|c3f64421973d:44
49,168,0,7a0d24d4d31c:54|97c3063d8125:56|c3f64421973d:58
50,179,0,7a0d24d4d31c:61|97c3063d8125:59|c3f64421973d:59
51,148,0,7a0d24d4d31c:50|97c3063d8125:50|c3f64421973d:48
52,128,0,7a0d24d4d31c:41|97c3063d8125:43|c3f64421973d:44
53,124,0,7a0d24d4d31c:42|97c3063d8125:41|c3f64421973d:41
54,247,0,7a0d24d4d31c:89|97c3063d8125:87|c3f64421973d:71
55,156,0,7a0d24d4d31c:77|97c3063d8125:79
56,159,0,7a0d24d4d31c:78|97c3063d8125:81
57,251,0,7a0d24d4d31c:128|97c3063d8125:123
58,179,0,7a0d24d4d31c:89|97c3063d8125:90
59,136,0,7a0d24d4d31c:67|97c3063d8125:69
60,215,0,7a0d24d4d31c:108|97c3063d8125:107
61,197,0,7a0d24d4d31c:99|97c3063d8125:98
62,183,0,7a0d24d4d31c:91|97c3063d8125:92
63,177,0,7a0d24d4d31c:89|97c3063d8125:88
64,320,0,7a0d24d4d31c:160|97c3063d8125:160
65,209,0,7a0d24d4d31c:104|97c3063d8125:105
66,140,0,7a0d24d4d31c:69|97c3063d8125:71
67,25,0,7a0d24d4d31c:14|97c3063d8125:11
1 second ok failed instances
2 0 324 0 907f96037a59:162|c3f64421973d:162
3 1 218 0 907f96037a59:108|c3f64421973d:110
4 2 188 0 907f96037a59:95|c3f64421973d:93
5 3 313 0 907f96037a59:156|c3f64421973d:157
6 4 309 0 907f96037a59:153|c3f64421973d:156
7 5 132 0 907f96037a59:67|c3f64421973d:65
8 6 224 0 907f96037a59:113|c3f64421973d:111
9 7 250 0 907f96037a59:125|c3f64421973d:125
10 8 176 0 907f96037a59:86|c3f64421973d:90
11 9 232 0 907f96037a59:118|c3f64421973d:114
12 10 337 0 907f96037a59:168|c3f64421973d:169
13 11 181 0 907f96037a59:89|c3f64421973d:92
14 12 227 0 907f96037a59:115|c3f64421973d:112
15 13 290 0 907f96037a59:145|c3f64421973d:145
16 14 277 0 907f96037a59:138|c3f64421973d:139
17 15 204 0 907f96037a59:103|c3f64421973d:101
18 16 302 0 907f96037a59:151|c3f64421973d:151
19 17 203 0 907f96037a59:100|c3f64421973d:103
20 18 108 0 907f96037a59:54|c3f64421973d:54
21 19 216 0 907f96037a59:110|c3f64421973d:106
22 20 252 0 907f96037a59:125|c3f64421973d:127
23 21 145 0 907f96037a59:72|c3f64421973d:73
24 22 214 0 907f96037a59:108|c3f64421973d:106
25 23 231 0 907f96037a59:115|c3f64421973d:116
26 24 187 0 907f96037a59:92|c3f64421973d:95
27 25 138 0 907f96037a59:69|c3f64421973d:69
28 26 237 0 907f96037a59:120|c3f64421973d:117
29 27 249 0 907f96037a59:125|c3f64421973d:124
30 28 188 0 907f96037a59:77|97c3063d8125:34|c3f64421973d:77
31 29 334 0 907f96037a59:111|97c3063d8125:111|c3f64421973d:112
32 30 289 0 907f96037a59:96|97c3063d8125:97|c3f64421973d:96
33 31 139 0 907f96037a59:45|97c3063d8125:47|c3f64421973d:47
34 32 232 0 907f96037a59:79|97c3063d8125:77|c3f64421973d:76
35 33 240 0 907f96037a59:80|97c3063d8125:79|c3f64421973d:81
36 34 170 0 907f96037a59:55|97c3063d8125:58|c3f64421973d:57
37 35 260 0 907f96037a59:29|97c3063d8125:116|c3f64421973d:115
38 36 334 0 97c3063d8125:166|c3f64421973d:168
39 37 203 0 97c3063d8125:103|c3f64421973d:100
40 38 176 0 97c3063d8125:87|c3f64421973d:89
41 39 232 0 97c3063d8125:116|c3f64421973d:116
42 40 182 0 97c3063d8125:92|c3f64421973d:90
43 41 88 0 97c3063d8125:44|c3f64421973d:44
44 42 202 0 97c3063d8125:100|c3f64421973d:102
45 43 225 0 97c3063d8125:112|c3f64421973d:113
46 44 171 0 97c3063d8125:87|c3f64421973d:84
47 45 197 0 97c3063d8125:97|c3f64421973d:100
48 46 342 0 7a0d24d4d31c:9|97c3063d8125:167|c3f64421973d:166
49 47 195 0 7a0d24d4d31c:65|97c3063d8125:65|c3f64421973d:65
50 48 134 0 7a0d24d4d31c:45|97c3063d8125:45|c3f64421973d:44
51 49 168 0 7a0d24d4d31c:54|97c3063d8125:56|c3f64421973d:58
52 50 179 0 7a0d24d4d31c:61|97c3063d8125:59|c3f64421973d:59
53 51 148 0 7a0d24d4d31c:50|97c3063d8125:50|c3f64421973d:48
54 52 128 0 7a0d24d4d31c:41|97c3063d8125:43|c3f64421973d:44
55 53 124 0 7a0d24d4d31c:42|97c3063d8125:41|c3f64421973d:41
56 54 247 0 7a0d24d4d31c:89|97c3063d8125:87|c3f64421973d:71
57 55 156 0 7a0d24d4d31c:77|97c3063d8125:79
58 56 159 0 7a0d24d4d31c:78|97c3063d8125:81
59 57 251 0 7a0d24d4d31c:128|97c3063d8125:123
60 58 179 0 7a0d24d4d31c:89|97c3063d8125:90
61 59 136 0 7a0d24d4d31c:67|97c3063d8125:69
62 60 215 0 7a0d24d4d31c:108|97c3063d8125:107
63 61 197 0 7a0d24d4d31c:99|97c3063d8125:98
64 62 183 0 7a0d24d4d31c:91|97c3063d8125:92
65 63 177 0 7a0d24d4d31c:89|97c3063d8125:88
66 64 320 0 7a0d24d4d31c:160|97c3063d8125:160
67 65 209 0 7a0d24d4d31c:104|97c3063d8125:105
68 66 140 0 7a0d24d4d31c:69|97c3063d8125:71
69 67 25 0 7a0d24d4d31c:14|97c3063d8125:11
+214
View File
@@ -0,0 +1,214 @@
# Own tests: unit + stand (pytest -v); failures restored in finally
# 2026-10-19T15:12:54Z, commit 2708732 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
$ python -m pytest student-tests -v
============================= test session starts =============================
platform win32 -- Python 3.14.6, pytest-9.1.1, pluggy-1.7.0 -- C:\Users\wfa22\Desktop\Высоконагруженные приложения\mtusi2026lab3-clean\.venv\Scripts\python.exe
rootdir: C:\Users\wfa22\Desktop\Высоконагруженные приложения\mtusi2026lab3-clean
plugins: anyio-4.15.1
collecting ... collected 201 items
student-tests/test_api.py::test_register_never_returns_password_and_login_returns_bearer PASSED [ 0%]
student-tests/test_api.py::test_duplicate_username_is_409 PASSED [ 0%]
student-tests/test_api.py::test_bad_password_on_registration_is_400[short-11ch] PASSED [ 1%]
student-tests/test_api.py::test_bad_password_on_registration_is_400[ ] PASSED [ 1%]
student-tests/test_api.py::test_bad_password_on_registration_is_400[ \t \n \t \n \t \n ] PASSED [ 2%]
student-tests/test_api.py::test_bad_password_on_registration_is_400[\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f] PASSED [ 2%]
student-tests/test_api.py::test_edge_passwords_register_and_login[\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f] PASSED [ 3%]
student-tests/test_api.py::test_edge_passwords_register_and_login[with:colons:inside] PASSED [ 3%]
student-tests/test_api.py::test_edge_passwords_register_and_login[Twelve chars] PASSED [ 4%]
student-tests/test_api.py::test_wrong_password_and_unknown_user_give_the_same_401 PASSED [ 4%]
student-tests/test_api.py::test_malformed_basic_is_401_not_500[Basic \xe9] PASSED [ 5%]
student-tests/test_api.py::test_malformed_basic_is_401_not_500[Basic !!!] PASSED [ 5%]
student-tests/test_api.py::test_malformed_basic_is_401_not_500[Basic bm9fY29sb24=] PASSED [ 6%]
student-tests/test_api.py::test_malformed_basic_is_401_not_500[Basic //46eA==] PASSED [ 6%]
student-tests/test_api.py::test_malformed_basic_is_401_not_500[Bearer aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa] PASSED [ 7%]
student-tests/test_api.py::test_malformed_basic_is_401_not_500[] PASSED [ 7%]
student-tests/test_api.py::test_logout_revokes_only_this_token PASSED [ 8%]
student-tests/test_api.py::test_x_user_id_basic_and_garbage_do_not_authenticate PASSED [ 8%]
student-tests/test_api.py::test_malformed_json_is_400[{] PASSED [ 9%]
student-tests/test_api.py::test_malformed_json_is_400[] PASSED [ 9%]
student-tests/test_api.py::test_malformed_json_is_400[[]] PASSED [ 10%]
student-tests/test_api.py::test_malformed_json_is_400[null] PASSED [ 10%]
student-tests/test_api.py::test_malformed_json_is_400[{"username": "abc_def", "display_name": NaN, "password": "Twelve-chars"}] PASSED [ 11%]
student-tests/test_api.py::test_malformed_json_is_400[{"username": "abc_def", "display_name": "x", "password": "Twelve-chars"}{"a": 1}] PASSED [ 11%]
student-tests/test_api.py::test_malformed_json_is_400[{"username": "abc_def", "display_name": "\ud800", "password": "Twelve-chars"}] PASSED [ 12%]
student-tests/test_api.py::test_invalid_registration_fields_are_400[body0] PASSED [ 12%]
student-tests/test_api.py::test_invalid_registration_fields_are_400[body1] PASSED [ 13%]
student-tests/test_api.py::test_invalid_registration_fields_are_400[body2] PASSED [ 13%]
student-tests/test_api.py::test_invalid_registration_fields_are_400[body3] PASSED [ 14%]
student-tests/test_api.py::test_invalid_registration_fields_are_400[body4] PASSED [ 14%]
student-tests/test_api.py::test_invalid_registration_fields_are_400[body5] PASSED [ 15%]
student-tests/test_api.py::test_unicode_limits_are_code_points_and_nul_is_stored PASSED [ 15%]
student-tests/test_api.py::test_oversized_body_and_wrong_content_type PASSED [ 16%]
student-tests/test_api.py::test_chat_validation PASSED [ 16%]
student-tests/test_api.py::test_foreign_chat_is_indistinguishable_from_missing PASSED [ 17%]
student-tests/test_api.py::test_chat_list_order_pagination_and_isolation PASSED [ 17%]
student-tests/test_api.py::test_history_pages_without_loss_or_duplicates PASSED [ 18%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[limit=0] PASSED [ 18%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[limit=101] PASSED [ 19%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[limit=abc] PASSED [ 19%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[limit=1&limit=2] PASSED [ 20%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[cursor=] PASSED [ 20%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[cursor=garbage] PASSED [ 21%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[cursor=djF8bXx4fDF8Mg] PASSED [ 21%]
student-tests/test_api.py::test_bad_limit_or_cursor_is_400[cursor=djF8bXx4fDk5OTk5OTk5OTk5OTk5OTk5OTk5fHk=] PASSED [ 22%]
student-tests/test_api.py::test_cursor_bound_to_chat PASSED [ 22%]
student-tests/test_api.py::test_read_after_write PASSED [ 23%]
student-tests/test_api.py::test_malformed_cursor_values_are_400_on_both_lists PASSED [ 23%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[\xff] PASSED [ 24%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"first","TEXT":"second"}] PASSED [ 24%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"a","text":"b"}] PASSED [ 25%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":123}] PASSED [ 25%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"valid"}}] PASSED [ 26%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"valid"}]] PASSED [ 26%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"x"} {}] PASSED [ 27%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"x","sender_id":"11111111-1111-4111-8111-111111111111"}] PASSED [ 27%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"x","attachment_ids":null}] PASSED [ 28%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":"x","attachment_ids":["nope"]}] PASSED [ 28%]
student-tests/test_api.py::test_bad_message_body_is_400_and_nothing_is_stored[{"text":" \n\t "}] PASSED [ 29%]
student-tests/test_api.py::test_ff_byte_body_on_registration_is_contract_error PASSED [ 29%]
student-tests/test_api.py::test_text_is_stored_verbatim_and_length_counts_raw_text PASSED [ 30%]
student-tests/test_api.py::test_unknown_attachment_ids_are_404 PASSED [ 30%]
student-tests/test_api.py::test_member_ids_are_compared_as_uuids_not_strings PASSED [ 31%]
student-tests/test_api.py::test_non_ascii_or_garbage_bearer_is_401_not_500[Bearer \xe9] PASSED [ 31%]
student-tests/test_api.py::test_non_ascii_or_garbage_bearer_is_401_not_500[Bearer .] PASSED [ 32%]
student-tests/test_api.py::test_non_ascii_or_garbage_bearer_is_401_not_500[Bearer a b] PASSED [ 32%]
student-tests/test_api.py::test_non_ascii_or_garbage_bearer_is_401_not_500[Bearer \xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9] PASSED [ 33%]
student-tests/test_api.py::test_non_ascii_or_garbage_bearer_is_401_not_500[bearer\taaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa] PASSED [ 33%]
student-tests/test_api.py::test_login_with_a_body_is_rejected PASSED [ 34%]
student-tests/test_api.py::test_password_range_is_12_to_128_code_points PASSED [ 34%]
student-tests/test_api.py::test_every_api_response_carries_instance_and_request_id PASSED [ 35%]
student-tests/test_api.py::test_unicode_whitespace_only_strings_are_400[\xa0] PASSED [ 35%]
student-tests/test_api.py::test_unicode_whitespace_only_strings_are_400[\u2003\u2003] PASSED [ 36%]
student-tests/test_api.py::test_unicode_whitespace_only_strings_are_400[ \xa0\t\u3000] PASSED [ 36%]
student-tests/test_api.py::test_odd_registration_bodies_are_400_and_the_api_survives[{"username":"num_probe","display_name":"x","password":1e1000}] PASSED [ 37%]
student-tests/test_api.py::test_odd_registration_bodies_are_400_and_the_api_survives[{"username":"ff_probe","display_name":"x","password":"12345678901\xff"}] PASSED [ 37%]
student-tests/test_api.py::test_odd_registration_bodies_are_400_and_the_api_survives[{"username":"num_probe","display_name":1,"password":"Twelve-chars-ok"}] PASSED [ 38%]
student-tests/test_api.py::test_more_than_100_members_is_400 PASSED [ 38%]
student-tests/test_api.py::test_cursor_of_one_list_is_rejected_by_the_other PASSED [ 39%]
student-tests/test_balancing.py::test_one_session_is_served_by_every_real_replica PASSED [ 39%]
student-tests/test_balancing.py::test_write_on_one_replica_is_read_on_the_other PASSED [ 40%]
student-tests/test_balancing.py::test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready[0-pg_hba] PASSED [ 40%]
student-tests/test_balancing.py::test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready[0-network] PASSED [ 41%]
student-tests/test_balancing.py::test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready[1-pg_hba] PASSED [ 41%]
student-tests/test_balancing.py::test_lb_stops_routing_to_a_replica_that_is_alive_but_not_ready[1-network] PASSED [ 42%]
student-tests/test_balancing.py::test_graceful_stop_of_each_replica_loses_no_request[0] PASSED [ 42%]
student-tests/test_balancing.py::test_graceful_stop_of_each_replica_loses_no_request[1] PASSED [ 43%]
student-tests/test_balancing.py::test_killed_replica_bounded_errors_recovery_and_no_duplicates[0] PASSED [ 43%]
student-tests/test_balancing.py::test_killed_replica_bounded_errors_recovery_and_no_duplicates[1] PASSED [ 44%]
student-tests/test_balancing.py::test_graceful_shutdown_order_in_logs PASSED [ 44%]
student-tests/test_balancing.py::test_full_restart_lb_rediscovers_all_replicas PASSED [ 45%]
student-tests/test_balancing.py::test_scale_up_and_down_is_picked_up_without_errors PASSED [ 45%]
student-tests/test_balancing.py::test_compose_policy_limits_ports_secrets_images PASSED [ 46%]
student-tests/test_balancing.py::test_running_containers_have_limits_and_no_bypass_ports PASSED [ 46%]
student-tests/test_balancing.py::test_migrations_ran_once_by_the_migrate_service PASSED [ 47%]
student-tests/test_db.py::test_registration_race_exactly_one_201_and_seven_409 PASSED [ 47%]
student-tests/test_db.py::test_create_chat_rolls_back_after_the_chat_row_was_inserted PASSED [ 48%]
student-tests/test_db.py::test_migrations_rerun_is_a_noop PASSED [ 48%]
student-tests/test_db.py::test_100_messages_from_10_clients_keep_order_and_cursor_walk PASSED [ 49%]
student-tests/test_db.py::test_secrets_are_not_stored_in_plain_text PASSED [ 49%]
student-tests/test_explain.py::test_late_page_uses_index_range_not_filter PASSED [ 50%]
student-tests/test_failures.py::test_valkey_failure[stop] PASSED [ 50%]
student-tests/test_failures.py::test_valkey_failure[pause] PASSED [ 51%]
student-tests/test_failures.py::test_postgres_failure[stop] PASSED [ 51%]
student-tests/test_failures.py::test_postgres_failure[pause] PASSED [ 52%]
student-tests/test_failures.py::test_locked_table_gives_bounded_503 PASSED [ 52%]
student-tests/test_failures.py::test_logout_during_valkey_outage_is_never_lost PASSED [ 53%]
student-tests/test_idempotency.py::test_replay_returns_the_same_201_and_message_from_every_replica PASSED [ 53%]
student-tests/test_idempotency.py::test_other_body_with_the_same_key_is_409_and_keeps_the_original PASSED [ 54%]
student-tests/test_idempotency.py::test_key_order_and_missing_attachment_ids_are_the_same_body PASSED [ 54%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[] PASSED [ 55%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa] PASSED [ 55%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[\u043a\u043b\u044e\u0447] PASSED [ 56%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[a b] PASSED [ 56%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[a/b] PASSED [ 57%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[a,b] PASSED [ 57%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[a\tb] PASSED [ 58%]
student-tests/test_idempotency.py::test_missing_or_invalid_key_is_400[*] PASSED [ 58%]
student-tests/test_idempotency.py::test_repeated_key_header_is_400 PASSED [ 59%]
student-tests/test_idempotency.py::test_longest_key_and_every_allowed_character PASSED [ 59%]
student-tests/test_idempotency.py::test_scope_is_user_plus_chat_plus_key PASSED [ 60%]
student-tests/test_idempotency.py::test_outsider_and_revoked_session_never_see_the_stored_result PASSED [ 60%]
student-tests/test_idempotency.py::test_concurrent_same_key_through_different_replicas_creates_exactly_one PASSED [ 61%]
student-tests/test_idempotency.py::test_concurrent_same_key_different_bodies_one_wins PASSED [ 61%]
student-tests/test_idempotency.py::test_failure_after_the_key_insert_rolls_everything_back PASSED [ 62%]
student-tests/test_idempotency.py::test_client_timeout_then_retry_with_the_same_key_gives_the_committed_message PASSED [ 62%]
student-tests/test_idempotency.py::test_lock_timeout_is_a_transient_503_that_is_not_recorded PASSED [ 63%]
student-tests/test_idempotency.py::test_replay_after_the_serving_replica_is_stopped PASSED [ 63%]
student-tests/test_sessions.py::test_short_ttl_expires[valkey] PASSED [ 64%]
student-tests/test_sessions.py::test_short_ttl_expires[postgres] PASSED [ 64%]
student-tests/test_sessions.py::test_restart_does_not_extend_ttl PASSED [ 65%]
student-tests/test_sessions.py::test_postgres_session_store PASSED [ 65%]
student-tests/unit/test_unit.py::test_valid_passwords[Twelve-chars] PASSED [ 66%]
student-tests/unit/test_unit.py::test_valid_passwords[\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f] PASSED [ 66%]
student-tests/unit/test_unit.py::test_valid_passwords[\u043f\u0430\u0440\u043e\u043b\u044c:\u0441:\u0434\u0432\u043e\u0435\u0442\u043e\u0447\u0438\u044f\u043c\u0438] PASSED [ 67%]
student-tests/unit/test_unit.py::test_valid_passwords[\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600\U0001f600] PASSED [ 67%]
student-tests/unit/test_unit.py::test_valid_passwords[a b c d e f g] PASSED [ 68%]
student-tests/unit/test_unit.py::test_invalid_passwords[short-11ch] PASSED [ 68%]
student-tests/unit/test_unit.py::test_invalid_passwords[\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f\u044f] PASSED [ 69%]
student-tests/unit/test_unit.py::test_invalid_passwords[ ] PASSED [ 69%]
student-tests/unit/test_unit.py::test_invalid_passwords[\t\n\u3000\t\n\u3000\t\n\u3000\t\n\u3000\t\n\u3000] PASSED [ 70%]
student-tests/unit/test_unit.py::test_invalid_passwords[\ud800\ud800\ud800\ud800\ud800\ud800\ud800\ud800\ud800\ud800\ud800\ud800] PASSED [ 70%]
student-tests/unit/test_unit.py::test_invalid_passwords[None] PASSED [ 71%]
student-tests/unit/test_unit.py::test_invalid_passwords[123456789012] PASSED [ 71%]
student-tests/unit/test_unit.py::test_text_may_contain_nul_but_not_lone_surrogates PASSED [ 72%]
student-tests/unit/test_unit.py::test_cursor_roundtrip_and_binding PASSED [ 72%]
student-tests/unit/test_unit.py::test_broken_cursors_are_400_not_500[v1|m|3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11|253402300800000000|11111111-1111-4111-8111-111111111111] PASSED [ 73%]
student-tests/unit/test_unit.py::test_broken_cursors_are_400_not_500[v1|m|3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11|9999999999999999999|11111111-1111-4111-8111-111111111111] PASSED [ 73%]
student-tests/unit/test_unit.py::test_broken_cursors_are_400_not_500[v1|m|3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11|-1|11111111-1111-4111-8111-111111111111] PASSED [ 74%]
student-tests/unit/test_unit.py::test_broken_cursors_are_400_not_500[v1|m|3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11|1|not-a-uuid] PASSED [ 74%]
student-tests/unit/test_unit.py::test_broken_cursors_are_400_not_500[v1|m|3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11|1|11111111-1111-4111-8111-11111111111G] PASSED [ 75%]
student-tests/unit/test_unit.py::test_broken_cursors_are_400_not_500[v2|m|3f2c7a5e-0b7d-4a35-9a43-8d1f0f3b2c11|1|11111111-1111-4111-8111-111111111111] PASSED [ 75%]
student-tests/unit/test_unit.py::test_time_helpers_are_exact PASSED [ 76%]
student-tests/unit/test_unit.py::test_basic_is_parsed_and_password_may_contain_colons PASSED [ 76%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[None] PASSED [ 77%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[] PASSED [ 77%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic] PASSED [ 78%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic \xe9] PASSED [ 78%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic !!!!] PASSED [ 79%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic bm8tY29sb24=] PASSED [ 79%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic //46cHc=] PASSED [ 80%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic OnBhc3N3b3Jk] PASSED [ 80%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic YWxpY2U6] PASSED [ 81%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic 0JDQu9C40YHQsDpwdw==] PASSED [ 81%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Bearer aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa] PASSED [ 82%]
student-tests/unit/test_unit.py::test_bad_basic_is_401_with_basic_challenge[Basic AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA] PASSED [ 82%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[None] PASSED [ 83%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[] PASSED [ 83%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[Bearer] PASSED [ 84%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[Bearer short] PASSED [ 84%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[Bearer aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa!] PASSED [ 85%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[Basic YWxpY2U6cGFzc3dvcmQtMTIz] PASSED [ 85%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[Token aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa] PASSED [ 86%]
student-tests/unit/test_unit.py::test_bad_bearer_is_401_with_bearer_challenge[Bearer \xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9\xe9] PASSED [ 86%]
student-tests/unit/test_unit.py::test_bearer_ok PASSED [ 87%]
student-tests/unit/test_unit.py::test_argon2id_hash_and_verify PASSED [ 87%]
student-tests/unit/test_unit.py::test_settings_read_secret_files_and_contract_names PASSED [ 88%]
student-tests/unit/test_unit.py::test_valid_idempotency_keys[a] PASSED [ 88%]
student-tests/unit/test_unit.py::test_valid_idempotency_keys[kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk] PASSED [ 89%]
student-tests/unit/test_unit.py::test_valid_idempotency_keys[A-Za-z0-9._:-] PASSED [ 89%]
student-tests/unit/test_unit.py::test_valid_idempotency_keys[2026-10-16T10:00:00.1] PASSED [ 90%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[None] PASSED [ 90%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[] PASSED [ 91%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk] PASSED [ 91%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[\u043a\u043b\u044e\u0447] PASSED [ 92%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[a b] PASSED [ 92%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[a/b] PASSED [ 93%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[a,b] PASSED [ 93%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[a\tb] PASSED [ 94%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[a\x00b] PASSED [ 94%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[*] PASSED [ 95%]
student-tests/unit/test_unit.py::test_invalid_idempotency_keys[\xe9] PASSED [ 95%]
student-tests/unit/test_unit.py::test_duplicate_keys_and_garbage_json_are_rejected[{"text":"a","text":"b"}] PASSED [ 96%]
student-tests/unit/test_unit.py::test_duplicate_keys_and_garbage_json_are_rejected[{"a":{"x":1,"x":2}}] PASSED [ 96%]
student-tests/unit/test_unit.py::test_duplicate_keys_and_garbage_json_are_rejected[\xff] PASSED [ 97%]
student-tests/unit/test_unit.py::test_duplicate_keys_and_garbage_json_are_rejected[{"text":"x"} {}] PASSED [ 97%]
student-tests/unit/test_unit.py::test_attachment_ids_absent_equals_empty_and_is_validated PASSED [ 98%]
student-tests/unit/test_unit.py::test_request_fingerprint_ignores_key_order_and_default_attachments PASSED [ 98%]
student-tests/unit/test_unit.py::test_instance_id_defaults_to_hostname_and_is_validated PASSED [ 99%]
student-tests/unit/test_unit.py::test_idempotency_ttl_cannot_be_below_24_hours PASSED [ 99%]
student-tests/unit/test_unit.py::test_message_write_path_setting PASSED [100%]
======================= 201 passed in 479.84s (0:07:59) =======================
[exit 0, 481.1 s]
+74
View File
@@ -0,0 +1,74 @@
# make test-failover with scripts/stop-one: which container is really stopped, restore, repeat
# 2026-10-19T15:12:01Z, commit 2708732 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
--- run 1; running API replicas BEFORE:
mtusi-lab3-api-1 host=94b38b941a8d running exit=0 pid=27148 started=2026-10-19T15:11:53.132
mtusi-lab3-api-2 host=a8d9cffd7a6e running exit=0 pid=27091 started=2026-10-19T15:11:52.824
$ make test-failover ACTION_SCRIPT=scripts/stop-one.cmd
python tests/resilience.py --mode failover --action-script "scripts/stop-one.cmd" --base-url "http://localhost:8080"
Executing explicit failover hook; hook output suppressed to avoid leaking secrets.
{"observed_instances_before": 2, "successful_writes": 131, "failed_attempts": 0, "first_recovery_seconds": 5.902}
PASS: failover preserved confirmed data and existing session. Hook scope needs operator evidence.
[exit 0, 16.4 s]
AFTER (the stopped replica, left stopped by the hook):
mtusi-lab3-api-1 host=94b38b941a8d exited exit=0 pid=0 started=2026-10-19T15:11:53.132
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py status
api2 172.23.0.2 UP mtusi-lab3-api-2 X-Instance-Id=a8d9cffd7a6e
(other slots: MAINT, no container)
[exit 0, 0.7 s]
$ sh scripts/restore-api.sh
mtusi-lab3-api-1 Up 6 seconds (healthy)
mtusi-lab3-api-2 Up 33 seconds (healthy)
Container mtusi-lab3-api-2 Running
Container mtusi-lab3-api-1 Starting
Container mtusi-lab3-api-1 Started
Container mtusi-lab3-api-2 Waiting
Container mtusi-lab3-api-1 Waiting
Container mtusi-lab3-api-2 Healthy
Container mtusi-lab3-api-1 Healthy
[exit 0, 6.9 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py wait-up 2
2 API replica(s) UP in HAProxy after 0.7 s: ['94b38b941a8d', 'a8d9cffd7a6e']
[exit 0, 1.4 s]
--- run 2; running API replicas BEFORE:
mtusi-lab3-api-1 host=94b38b941a8d running exit=0 pid=28373 started=2026-10-19T15:12:20.518
mtusi-lab3-api-2 host=a8d9cffd7a6e running exit=0 pid=27091 started=2026-10-19T15:11:52.824
$ make test-failover ACTION_SCRIPT=scripts/stop-one.cmd
python tests/resilience.py --mode failover --action-script "scripts/stop-one.cmd" --base-url "http://localhost:8080"
Executing explicit failover hook; hook output suppressed to avoid leaking secrets.
{"observed_instances_before": 2, "successful_writes": 132, "failed_attempts": 0, "first_recovery_seconds": 5.667}
PASS: failover preserved confirmed data and existing session. Hook scope needs operator evidence.
[exit 0, 16.3 s]
AFTER (the stopped replica, left stopped by the hook):
mtusi-lab3-api-2 host=a8d9cffd7a6e exited exit=0 pid=0 started=2026-10-19T15:11:52.824
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py status
api1 172.23.0.3 UP mtusi-lab3-api-1 X-Instance-Id=94b38b941a8d
(other slots: MAINT, no container)
[exit 0, 0.9 s]
$ sh scripts/restore-api.sh
mtusi-lab3-api-1 Up 32 seconds (healthy)
mtusi-lab3-api-2 Up 6 seconds (healthy)
Container mtusi-lab3-api-1 Running
Container mtusi-lab3-api-2 Starting
Container mtusi-lab3-api-2 Started
Container mtusi-lab3-api-2 Waiting
Container mtusi-lab3-api-1 Waiting
Container mtusi-lab3-api-2 Healthy
Container mtusi-lab3-api-1 Healthy
[exit 0, 7.0 s]
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py wait-up 2
2 API replica(s) UP in HAProxy after 0.7 s: ['94b38b941a8d', 'a8d9cffd7a6e']
[exit 0, 1.6 s]
Replicas stopped by the public test across runs: ['mtusi-lab3-api-1', 'mtusi-lab3-api-2']
+28
View File
@@ -0,0 +1,28 @@
# make test-persistence: restart of LB, all API replicas, PostgreSQL, Valkey
# 2026-10-19T15:11:37Z, commit 2708732 fix: LB и на [::1] — localhost на Windows стоил 2 с на запрос; сдвиг цели failover
Containers BEFORE:
mtusi-lab3-api-1 host=94b38b941a8d running exit=0 pid=23944 started=2026-10-19T15:10:34.257
mtusi-lab3-api-2 host=a8d9cffd7a6e running exit=0 pid=23887 started=2026-10-19T15:10:33.970
mtusi-lab3-lb-1 host=d903237b9e99 running exit=0 pid=24172 started=2026-10-19T15:10:40.108
mtusi-lab3-migrate-1 host=d39cf5a2265b exited exit=0 pid=0 started=2026-10-19T15:10:31.699
mtusi-lab3-postgres-1 host=3ee559fb5f7f running exit=0 pid=23533 started=2026-10-19T15:10:25.911
mtusi-lab3-valkey-1 host=8594c96d36f9 running exit=0 pid=23621 started=2026-10-19T15:10:26.220
$ make test-persistence ACTION_SCRIPT=scripts/restart.cmd
python tests/resilience.py --mode persistence --action-script "scripts/restart.cmd" --base-url "http://localhost:8080"
Executing explicit persistence hook; hook output suppressed to avoid leaking secrets.
PASS: persistence preserved confirmed data and existing session. Hook scope needs operator evidence.
[exit 0, 22.6 s]
Containers AFTER (new pid/started for lb, api-*, postgres, valkey; volumes kept):
mtusi-lab3-api-1 host=94b38b941a8d running exit=0 pid=27148 started=2026-10-19T15:11:53.132
mtusi-lab3-api-2 host=a8d9cffd7a6e running exit=0 pid=27091 started=2026-10-19T15:11:52.824
mtusi-lab3-lb-1 host=d903237b9e99 running exit=0 pid=27452 started=2026-10-19T15:11:59.277
mtusi-lab3-migrate-1 host=d39cf5a2265b exited exit=0 pid=0 started=2026-10-19T15:11:54.449
mtusi-lab3-postgres-1 host=3ee559fb5f7f running exit=0 pid=26819 started=2026-10-19T15:11:44.558
mtusi-lab3-valkey-1 host=8594c96d36f9 running exit=0 pid=26779 started=2026-10-19T15:11:44.461
$ C:\Users\wfa22\AppData\Local\Python\pythoncore-3.14-64\python.exe scripts/lbctl.py wait-up 2
2 API replica(s) UP in HAProxy after 0.6 s: ['94b38b941a8d', 'a8d9cffd7a6e']
[exit 0, 1.4 s]
@@ -0,0 +1,65 @@
{
"started": "2026-10-19T11:02:28+00:00",
"measured_commit": "3f45eb1",
"image": "mtusi-messenger:bench-3f45eb1",
"bench_tooling_dirty": false,
"host": {
"os": "Windows 10",
"docker": "29.8.0 8 CPU 15686365184 B 6.18.54.1-microsoft-standard-WSL2"
},
"cpusets": "k6 0-1, lb+valkey 2, postgres 3-4, api 5-7 (bench/track1/compose.bench.yaml)",
"workload": "k6 constant-arrival-rate POST message, new Idempotency-Key, 64 chats, warmup 10s + measure 20s per step, fresh stand per config x repeat",
"rates": [
100,
150,
200,
250,
300,
350
],
"repeats": 3,
"configs": [
{
"replicas": 1,
"api_cpus": 2.0,
"api_memory": "768m",
"pg_cpus": 2.0,
"pool": 10,
"env": {}
},
{
"replicas": 2,
"api_cpus": 1.0,
"api_memory": "384m",
"pg_cpus": 2.0,
"pool": 10,
"env": {}
},
{
"replicas": 4,
"api_cpus": 0.5,
"api_memory": "192m",
"pg_cpus": 2.0,
"pool": 10,
"env": {}
}
],
"args": [
"--git-ref",
"3f45eb1",
"--repeats",
"3",
"--series",
"A-fixed-total",
"--label",
"baseline",
"--rates",
"100,150,200,250,300,350",
"--config",
"replicas=1,api_cpus=2.0,api_memory=768m",
"--config",
"replicas=2,api_cpus=1.0,api_memory=384m",
"--config",
"replicas=4,api_cpus=0.5,api_memory=192m"
]
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2000,
"ok": 2000,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 7.5202495,
"ok_p95_ms": 28.052189000000002,
"ok_p99_ms": 98.54706085999999,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 75.4,
"cpu_max": 119.3,
"samples": 9
},
"api-1": {
"cpu_median": 75.4,
"cpu_max": 119.3,
"mem_last": "69.9MiB",
"samples": 9
},
"k6": {
"cpu_median": 12.2,
"cpu_max": 14.4,
"mem_last": "25.89MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 4.3,
"cpu_max": 18.6,
"mem_last": "11.44MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 25.6,
"cpu_max": 36.5,
"mem_last": "72.65MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 1.6,
"cpu_max": 6.5,
"mem_last": "4.047MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 4034,
"pool_wait_ms_avg": 0.008,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 2.935,
"api_cpu_s": 10.62,
"intervals": 10
},
"pg": {
"xact_commit": 12109,
"xact_rollback": 0,
"tup_inserted": 6113,
"blks_read": 60,
"blks_hit": 146164,
"wal_records": 43185,
"wal_bytes": 4257360
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3001,
"total_ms": 709.9,
"mean_ms": 0.2366,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3001,
"total_ms": 300.5,
"mean_ms": 0.1001,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6002,
"total_ms": 211.6,
"mean_ms": 0.0353,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3001,
"total_ms": 184.2,
"mean_ms": 0.0614,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 3001,
"total_ms": 58.0,
"mean_ms": 0.0193,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6048,
"total_ms": 8.1,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 6048,
"total_ms": 7.6,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 46,
"total_ms": 0.3,
"mean_ms": 0.0055,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1188,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0367,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.017,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.05,
"statements_per_ok": 15.08,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2992,
"ok": 2992,
"dropped": 9,
"ok_rps": 149.6,
"error_ratio": 0,
"ok_p50_ms": 8.871513,
"ok_p95_ms": 118.67201719999996,
"ok_p99_ms": 171.77328262000003,
"vus_max": 55,
"cpu": {
"api_total": {
"cpu_median": 78.6,
"cpu_max": 142.5,
"samples": 10
},
"api-1": {
"cpu_median": 78.6,
"cpu_max": 142.5,
"mem_last": "56.41MiB",
"samples": 10
},
"k6": {
"cpu_median": 17.3,
"cpu_max": 21.8,
"mem_last": "29.5MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 5.8,
"cpu_max": 14.0,
"mem_last": "11.67MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 27.4,
"cpu_max": 41.8,
"mem_last": "76.07MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.8,
"cpu_max": 2.1,
"mem_last": "3.828MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6007,
"pool_wait_ms_avg": 1.152,
"pool_waiting_max": 8,
"pool_usage_ms_avg": 4.813,
"api_cpu_s": 14.27,
"intervals": 10
},
"pg": {
"xact_commit": 17853,
"xact_rollback": 0,
"tup_inserted": 8852,
"blks_read": 0,
"blks_hit": 206541,
"wal_records": 62680,
"wal_bytes": 5541231
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4487,
"total_ms": 1012.5,
"mean_ms": 0.2256,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4487,
"total_ms": 431.0,
"mean_ms": 0.0961,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 8974,
"total_ms": 323.6,
"mean_ms": 0.0361,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4487,
"total_ms": 268.1,
"mean_ms": 0.0598,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 4487,
"total_ms": 90.9,
"mean_ms": 0.0203,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9021,
"total_ms": 11.3,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 9021,
"total_ms": 11.2,
"mean_ms": 0.0012,
"query": "BEGIN"
},
{
"calls": 47,
"total_ms": 0.2,
"mean_ms": 0.0049,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.1575,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0436,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0228,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.97,
"statements_per_ok": 15.05,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3973,
"ok": 3973,
"dropped": 27,
"ok_rps": 198.7,
"error_ratio": 0,
"ok_p50_ms": 13.108659,
"ok_p95_ms": 141.66078979999992,
"ok_p99_ms": 229.97273719999976,
"vus_max": 82,
"cpu": {
"api_total": {
"cpu_median": 103.1,
"cpu_max": 138.9,
"samples": 10
},
"api-1": {
"cpu_median": 103.1,
"cpu_max": 138.9,
"mem_last": "69.35MiB",
"samples": 10
},
"k6": {
"cpu_median": 21.8,
"cpu_max": 25.2,
"mem_last": "40.04MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.6,
"cpu_max": 18.2,
"mem_last": "11.95MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 38.4,
"cpu_max": 45.6,
"mem_last": "79.4MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 3.7,
"cpu_max": 6.4,
"mem_last": "3.898MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7981,
"pool_wait_ms_avg": 5.268,
"pool_waiting_max": 22,
"pool_usage_ms_avg": 8.248,
"api_cpu_s": 18.57,
"intervals": 10
},
"pg": {
"xact_commit": 23747,
"xact_rollback": 0,
"tup_inserted": 11830,
"blks_read": 0,
"blks_hit": 284364,
"wal_records": 83562,
"wal_bytes": 7400028
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5958,
"total_ms": 1354.6,
"mean_ms": 0.2274,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5958,
"total_ms": 650.1,
"mean_ms": 0.1091,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5958,
"total_ms": 407.0,
"mean_ms": 0.0683,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11916,
"total_ms": 396.9,
"mean_ms": 0.0333,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5958,
"total_ms": 140.6,
"mean_ms": 0.0236,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11962,
"total_ms": 14.4,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 11962,
"total_ms": 13.9,
"mean_ms": 0.0012,
"query": "BEGIN"
},
{
"calls": 46,
"total_ms": 0.2,
"mean_ms": 0.0041,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0926,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0375,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0198,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.98,
"statements_per_ok": 15.03,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4180,
"ok": 4180,
"dropped": 820,
"ok_rps": 209.0,
"error_ratio": 0,
"ok_p50_ms": 2151.489912,
"ok_p95_ms": 2612.4429757,
"ok_p99_ms": 2703.39118036,
"vus_max": 769,
"cpu": {
"api_total": {
"cpu_median": 110.6,
"cpu_max": 142.7,
"samples": 10
},
"api-1": {
"cpu_median": 110.6,
"cpu_max": 142.7,
"mem_last": "76.81MiB",
"samples": 10
},
"k6": {
"cpu_median": 28.9,
"cpu_max": 56.1,
"mem_last": "253.6MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.6,
"cpu_max": 20.1,
"mem_last": "18.7MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.1,
"cpu_max": 42.0,
"mem_last": "84.11MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 5.9,
"cpu_max": 7.6,
"mem_last": "3.75MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7978,
"pool_wait_ms_avg": 469.808,
"pool_waiting_max": 189,
"pool_usage_ms_avg": 18.023,
"api_cpu_s": 20.02,
"intervals": 10
},
"pg": {
"xact_commit": 25327,
"xact_rollback": 0,
"tup_inserted": 12524,
"blks_read": 0,
"blks_hit": 310088,
"wal_records": 88730,
"wal_bytes": 7783781
},
"lb": {
"chkdown_delta": 0,
"qmax": 300,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6405,
"total_ms": 1494.9,
"mean_ms": 0.2334,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6405,
"total_ms": 723.2,
"mean_ms": 0.1129,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6405,
"total_ms": 462.9,
"mean_ms": 0.0723,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 12810,
"total_ms": 400.5,
"mean_ms": 0.0313,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6405,
"total_ms": 167.6,
"mean_ms": 0.0262,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 12849,
"total_ms": 16.0,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 12849,
"total_ms": 13.8,
"mean_ms": 0.0011,
"query": "BEGIN"
},
{
"calls": 39,
"total_ms": 0.2,
"mean_ms": 0.0042,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1284,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0413,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0213,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.06,
"statements_per_ok": 15.35,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 4086,
"ok": 4086,
"dropped": 1914,
"ok_rps": 204.3,
"error_ratio": 0,
"ok_p50_ms": 2793.8822490000002,
"ok_p95_ms": 3124.0045007500003,
"ok_p99_ms": 3211.33611675,
"vus_max": 1131,
"cpu": {
"api_total": {
"cpu_median": 99.0,
"cpu_max": 140.4,
"samples": 10
},
"api-1": {
"cpu_median": 99.0,
"cpu_max": 140.4,
"mem_last": "66MiB",
"samples": 10
},
"k6": {
"cpu_median": 27.4,
"cpu_max": 37.6,
"mem_last": "335.4MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.6,
"cpu_max": 19.2,
"mem_last": "21.18MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 46.8,
"cpu_max": 50.8,
"mem_last": "88.58MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.1,
"cpu_max": 2.2,
"mem_last": "3.688MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8073,
"pool_wait_ms_avg": 461.578,
"pool_waiting_max": 190,
"pool_usage_ms_avg": 18.144,
"api_cpu_s": 19.61,
"intervals": 10
},
"pg": {
"xact_commit": 26759,
"xact_rollback": 0,
"tup_inserted": 13324,
"blks_read": 0,
"blks_hit": 340810,
"wal_records": 93665,
"wal_bytes": 8654533
},
"lb": {
"chkdown_delta": 0,
"qmax": 400,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 31,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6567,
"total_ms": 1511.8,
"mean_ms": 0.2302,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6567,
"total_ms": 754.0,
"mean_ms": 0.1148,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6567,
"total_ms": 480.1,
"mean_ms": 0.0731,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 13134,
"total_ms": 401.4,
"mean_ms": 0.0306,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6567,
"total_ms": 166.0,
"mean_ms": 0.0253,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13172,
"total_ms": 15.5,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 13172,
"total_ms": 13.6,
"mean_ms": 0.001,
"query": "BEGIN"
},
{
"calls": 38,
"total_ms": 0.1,
"mean_ms": 0.0039,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0964,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0441,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0211,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.55,
"statements_per_ok": 16.1,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 3920,
"ok": 3920,
"dropped": 3081,
"ok_rps": 196.0,
"error_ratio": 0,
"ok_p50_ms": 2988.8524905,
"ok_p95_ms": 3335.9800228999998,
"ok_p99_ms": 3396.08057728,
"vus_max": 1200,
"cpu": {
"api_total": {
"cpu_median": 100.7,
"cpu_max": 148.8,
"samples": 10
},
"api-1": {
"cpu_median": 100.7,
"cpu_max": 148.8,
"mem_last": "79.36MiB",
"samples": 10
},
"k6": {
"cpu_median": 27.4,
"cpu_max": 52.8,
"mem_last": "400.3MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 8.9,
"cpu_max": 18.3,
"mem_last": "21.95MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 39.7,
"cpu_max": 52.0,
"mem_last": "92.44MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 3.0,
"cpu_max": 6.5,
"mem_last": "3.809MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7990,
"pool_wait_ms_avg": 469.138,
"pool_waiting_max": 187,
"pool_usage_ms_avg": 18.21,
"api_cpu_s": 19.95,
"intervals": 10
},
"pg": {
"xact_commit": 25748,
"xact_rollback": 0,
"tup_inserted": 12834,
"blks_read": 0,
"blks_hit": 333218,
"wal_records": 90519,
"wal_bytes": 7797623
},
"lb": {
"chkdown_delta": 0,
"qmax": 418,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 30,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6410,
"total_ms": 1503.6,
"mean_ms": 0.2346,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6410,
"total_ms": 736.8,
"mean_ms": 0.1149,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6410,
"total_ms": 474.8,
"mean_ms": 0.0741,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 12820,
"total_ms": 402.0,
"mean_ms": 0.0314,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6410,
"total_ms": 167.3,
"mean_ms": 0.0261,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 12859,
"total_ms": 15.5,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 12859,
"total_ms": 13.7,
"mean_ms": 0.0011,
"query": "BEGIN"
},
{
"calls": 39,
"total_ms": 0.2,
"mean_ms": 0.0048,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1017,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0445,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0178,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.57,
"statements_per_ok": 16.38,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2001,
"ok": 2001,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 7.250932,
"ok_p95_ms": 15.74861,
"ok_p99_ms": 37.804146,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 69.4,
"cpu_max": 102.8,
"samples": 9
},
"api-1": {
"cpu_median": 69.4,
"cpu_max": 102.8,
"mem_last": "57.66MiB",
"samples": 9
},
"k6": {
"cpu_median": 11.9,
"cpu_max": 15.4,
"mem_last": "25.71MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 4.3,
"cpu_max": 11.5,
"mem_last": "11.67MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 26.2,
"cpu_max": 32.2,
"mem_last": "61.41MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 1.4,
"cpu_max": 1.9,
"mem_last": "3.934MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 4029,
"pool_wait_ms_avg": 0.004,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 2.151,
"api_cpu_s": 9.53,
"intervals": 10
},
"pg": {
"xact_commit": 11936,
"xact_rollback": 0,
"tup_inserted": 6001,
"blks_read": 82,
"blks_hit": 140615,
"wal_records": 42648,
"wal_bytes": 4237390
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3002,
"total_ms": 649.2,
"mean_ms": 0.2163,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3002,
"total_ms": 280.2,
"mean_ms": 0.0933,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6004,
"total_ms": 205.2,
"mean_ms": 0.0342,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3002,
"total_ms": 169.3,
"mean_ms": 0.0564,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 3002,
"total_ms": 53.8,
"mean_ms": 0.0179,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6050,
"total_ms": 7.6,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 6050,
"total_ms": 7.2,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 46,
"total_ms": 0.2,
"mean_ms": 0.0045,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1391,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0659,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0269,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.97,
"statements_per_ok": 15.07,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2995,
"ok": 2995,
"dropped": 6,
"ok_rps": 149.8,
"error_ratio": 0,
"ok_p50_ms": 8.864475,
"ok_p95_ms": 65.56181429999967,
"ok_p99_ms": 149.6798073,
"vus_max": 53,
"cpu": {
"api_total": {
"cpu_median": 87.1,
"cpu_max": 110.5,
"samples": 10
},
"api-1": {
"cpu_median": 87.1,
"cpu_max": 110.5,
"mem_last": "57.74MiB",
"samples": 10
},
"k6": {
"cpu_median": 17.5,
"cpu_max": 19.2,
"mem_last": "28.53MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 5.7,
"cpu_max": 15.8,
"mem_last": "11.61MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 26.0,
"cpu_max": 30.4,
"mem_last": "74.88MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.8,
"cpu_max": 2.7,
"mem_last": "3.672MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6018,
"pool_wait_ms_avg": 0.601,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 4.125,
"api_cpu_s": 14.18,
"intervals": 10
},
"pg": {
"xact_commit": 17926,
"xact_rollback": 0,
"tup_inserted": 8918,
"blks_read": 0,
"blks_hit": 211242,
"wal_records": 63348,
"wal_bytes": 5575852
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4489,
"total_ms": 1027.3,
"mean_ms": 0.2289,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4489,
"total_ms": 430.8,
"mean_ms": 0.096,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 8978,
"total_ms": 323.4,
"mean_ms": 0.036,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4489,
"total_ms": 264.1,
"mean_ms": 0.0588,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 4489,
"total_ms": 88.6,
"mean_ms": 0.0197,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9026,
"total_ms": 11.5,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 9026,
"total_ms": 10.8,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 48,
"total_ms": 0.2,
"mean_ms": 0.0048,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1231,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0495,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0257,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.99,
"statements_per_ok": 15.04,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3970,
"ok": 3970,
"dropped": 30,
"ok_rps": 198.5,
"error_ratio": 0,
"ok_p50_ms": 22.8694485,
"ok_p95_ms": 200.6632921,
"ok_p99_ms": 247.88514344999996,
"vus_max": 83,
"cpu": {
"api_total": {
"cpu_median": 101.6,
"cpu_max": 133.6,
"samples": 10
},
"api-1": {
"cpu_median": 101.6,
"cpu_max": 133.6,
"mem_last": "58.41MiB",
"samples": 10
},
"k6": {
"cpu_median": 23.0,
"cpu_max": 25.5,
"mem_last": "41.43MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.8,
"cpu_max": 18.6,
"mem_last": "11.86MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 45.4,
"cpu_max": 48.8,
"mem_last": "78.33MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 6.0,
"cpu_max": 6.9,
"mem_last": "3.887MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7979,
"pool_wait_ms_avg": 9.836,
"pool_waiting_max": 27,
"pool_usage_ms_avg": 9.36,
"api_cpu_s": 18.76,
"intervals": 10
},
"pg": {
"xact_commit": 23607,
"xact_rollback": 0,
"tup_inserted": 11762,
"blks_read": 0,
"blks_hit": 282303,
"wal_records": 83100,
"wal_bytes": 7402176
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5956,
"total_ms": 1373.0,
"mean_ms": 0.2305,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5956,
"total_ms": 654.3,
"mean_ms": 0.1099,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5956,
"total_ms": 408.8,
"mean_ms": 0.0686,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11912,
"total_ms": 403.2,
"mean_ms": 0.0338,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5956,
"total_ms": 144.1,
"mean_ms": 0.0242,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11958,
"total_ms": 14.4,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 11958,
"total_ms": 13.8,
"mean_ms": 0.0012,
"query": "BEGIN"
},
{
"calls": 46,
"total_ms": 0.2,
"mean_ms": 0.0046,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0597,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0984,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0289,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.95,
"statements_per_ok": 15.04,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4278,
"ok": 4278,
"dropped": 723,
"ok_rps": 213.9,
"error_ratio": 0,
"ok_p50_ms": 2058.8974980000003,
"ok_p95_ms": 2484.172225,
"ok_p99_ms": 2522.18200123,
"vus_max": 790,
"cpu": {
"api_total": {
"cpu_median": 99.4,
"cpu_max": 146.0,
"samples": 10
},
"api-1": {
"cpu_median": 99.4,
"cpu_max": 146.0,
"mem_last": "64.08MiB",
"samples": 10
},
"k6": {
"cpu_median": 28.7,
"cpu_max": 47.6,
"mem_last": "262.8MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.4,
"cpu_max": 20.0,
"mem_last": "18.81MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 49.1,
"cpu_max": 71.9,
"mem_last": "82.6MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.2,
"cpu_max": 6.0,
"mem_last": "3.652MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8165,
"pool_wait_ms_avg": 456.896,
"pool_waiting_max": 186,
"pool_usage_ms_avg": 17.685,
"api_cpu_s": 20.03,
"intervals": 10
},
"pg": {
"xact_commit": 25970,
"xact_rollback": 0,
"tup_inserted": 12858,
"blks_read": 0,
"blks_hit": 318554,
"wal_records": 91214,
"wal_bytes": 7941124
},
"lb": {
"chkdown_delta": 0,
"qmax": 300,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6504,
"total_ms": 1527.9,
"mean_ms": 0.2349,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6504,
"total_ms": 736.2,
"mean_ms": 0.1132,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6504,
"total_ms": 464.0,
"mean_ms": 0.0713,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 13008,
"total_ms": 412.1,
"mean_ms": 0.0317,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6504,
"total_ms": 169.4,
"mean_ms": 0.026,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13047,
"total_ms": 14.6,
"mean_ms": 0.0011,
"query": "COMMIT"
},
{
"calls": 13047,
"total_ms": 13.3,
"mean_ms": 0.001,
"query": "BEGIN"
},
{
"calls": 39,
"total_ms": 0.2,
"mean_ms": 0.0043,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1218,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.046,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0278,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.07,
"statements_per_ok": 15.23,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 4007,
"ok": 4007,
"dropped": 1994,
"ok_rps": 200.3,
"error_ratio": 0,
"ok_p50_ms": 2824.94653,
"ok_p95_ms": 3517.4384534,
"ok_p99_ms": 3643.0758680599997,
"vus_max": 1150,
"cpu": {
"api_total": {
"cpu_median": 96.6,
"cpu_max": 141.8,
"samples": 10
},
"api-1": {
"cpu_median": 96.6,
"cpu_max": 141.8,
"mem_last": "66.42MiB",
"samples": 10
},
"k6": {
"cpu_median": 26.5,
"cpu_max": 59.7,
"mem_last": "341MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 8.9,
"cpu_max": 18.1,
"mem_last": "20.92MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 44.6,
"cpu_max": 50.0,
"mem_last": "87.11MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.2,
"cpu_max": 2.3,
"mem_last": "3.832MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8123,
"pool_wait_ms_avg": 460.307,
"pool_waiting_max": 188,
"pool_usage_ms_avg": 18.286,
"api_cpu_s": 19.75,
"intervals": 10
},
"pg": {
"xact_commit": 26113,
"xact_rollback": 0,
"tup_inserted": 13028,
"blks_read": 0,
"blks_hit": 332926,
"wal_records": 91844,
"wal_bytes": 8578256
},
"lb": {
"chkdown_delta": 0,
"qmax": 400,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6457,
"total_ms": 1490.1,
"mean_ms": 0.2308,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6457,
"total_ms": 722.1,
"mean_ms": 0.1118,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6457,
"total_ms": 456.1,
"mean_ms": 0.0706,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 12914,
"total_ms": 409.9,
"mean_ms": 0.0317,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6457,
"total_ms": 160.5,
"mean_ms": 0.0249,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 12952,
"total_ms": 14.3,
"mean_ms": 0.0011,
"query": "COMMIT"
},
{
"calls": 12952,
"total_ms": 13.3,
"mean_ms": 0.001,
"query": "BEGIN"
},
{
"calls": 38,
"total_ms": 0.2,
"mean_ms": 0.0049,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1171,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0395,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0362,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
}
],
"pg_commits_per_ok": 6.52,
"statements_per_ok": 16.14,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 4021,
"ok": 4021,
"dropped": 2979,
"ok_rps": 201.1,
"error_ratio": 0,
"ok_p50_ms": 2877.468829,
"ok_p95_ms": 3297.091424,
"ok_p99_ms": 3325.204687,
"vus_max": 1200,
"cpu": {
"api_total": {
"cpu_median": 97.7,
"cpu_max": 144.0,
"samples": 10
},
"api-1": {
"cpu_median": 97.7,
"cpu_max": 144.0,
"mem_last": "66.6MiB",
"samples": 10
},
"k6": {
"cpu_median": 27.6,
"cpu_max": 76.2,
"mem_last": "380MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.3,
"cpu_max": 19.1,
"mem_last": "21.74MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.9,
"cpu_max": 46.6,
"mem_last": "90.81MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.2,
"cpu_max": 5.6,
"mem_last": "3.695MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8122,
"pool_wait_ms_avg": 460.264,
"pool_waiting_max": 186,
"pool_usage_ms_avg": 17.89,
"api_cpu_s": 20.01,
"intervals": 10
},
"pg": {
"xact_commit": 26027,
"xact_rollback": 0,
"tup_inserted": 12960,
"blks_read": 0,
"blks_hit": 335958,
"wal_records": 91412,
"wal_bytes": 7815846
},
"lb": {
"chkdown_delta": 0,
"qmax": 416,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 27,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6487,
"total_ms": 1513.8,
"mean_ms": 0.2334,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6487,
"total_ms": 739.9,
"mean_ms": 0.1141,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6487,
"total_ms": 480.7,
"mean_ms": 0.0741,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 12974,
"total_ms": 410.6,
"mean_ms": 0.0317,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6487,
"total_ms": 170.5,
"mean_ms": 0.0263,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13012,
"total_ms": 14.6,
"mean_ms": 0.0011,
"query": "COMMIT"
},
{
"calls": 13012,
"total_ms": 13.7,
"mean_ms": 0.0011,
"query": "BEGIN"
},
{
"calls": 38,
"total_ms": 0.2,
"mean_ms": 0.0045,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.2,
"mean_ms": 0.0821,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1093,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0385,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.47,
"statements_per_ok": 16.16,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2001,
"ok": 2001,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 7.97685,
"ok_p95_ms": 15.575488,
"ok_p99_ms": 28.970835,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 62.3,
"cpu_max": 99.6,
"samples": 10
},
"api-1": {
"cpu_median": 62.3,
"cpu_max": 99.6,
"mem_last": "57.85MiB",
"samples": 10
},
"k6": {
"cpu_median": 11.7,
"cpu_max": 13.6,
"mem_last": "25.88MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 4.3,
"cpu_max": 12.6,
"mem_last": "11.48MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 26.8,
"cpu_max": 36.8,
"mem_last": "60.41MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.4,
"cpu_max": 5.6,
"mem_last": "3.812MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 4030,
"pool_wait_ms_avg": 0.0,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 2.397,
"api_cpu_s": 9.69,
"intervals": 10
},
"pg": {
"xact_commit": 12018,
"xact_rollback": 0,
"tup_inserted": 6013,
"blks_read": 60,
"blks_hit": 141178,
"wal_records": 42996,
"wal_bytes": 4221770
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3002,
"total_ms": 655.0,
"mean_ms": 0.2182,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3002,
"total_ms": 272.2,
"mean_ms": 0.0907,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6004,
"total_ms": 204.9,
"mean_ms": 0.0341,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3002,
"total_ms": 169.8,
"mean_ms": 0.0566,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 3002,
"total_ms": 54.4,
"mean_ms": 0.0181,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6052,
"total_ms": 7.4,
"mean_ms": 0.0012,
"query": "BEGIN"
},
{
"calls": 6052,
"total_ms": 6.9,
"mean_ms": 0.0011,
"query": "COMMIT"
},
{
"calls": 48,
"total_ms": 0.2,
"mean_ms": 0.0048,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.1525,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0609,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0238,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.01,
"statements_per_ok": 15.08,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2997,
"ok": 2997,
"dropped": 4,
"ok_rps": 149.8,
"error_ratio": 0,
"ok_p50_ms": 8.971883,
"ok_p95_ms": 64.57286139999982,
"ok_p99_ms": 112.78668699999997,
"vus_max": 49,
"cpu": {
"api_total": {
"cpu_median": 83.8,
"cpu_max": 130.1,
"samples": 10
},
"api-1": {
"cpu_median": 83.8,
"cpu_max": 130.1,
"mem_last": "57.81MiB",
"samples": 10
},
"k6": {
"cpu_median": 17.7,
"cpu_max": 20.6,
"mem_last": "29.3MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 5.9,
"cpu_max": 19.2,
"mem_last": "11.62MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 34.0,
"cpu_max": 43.5,
"mem_last": "75.33MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.8,
"cpu_max": 2.2,
"mem_last": "4.129MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6034,
"pool_wait_ms_avg": 0.674,
"pool_waiting_max": 2,
"pool_usage_ms_avg": 4.393,
"api_cpu_s": 14.25,
"intervals": 10
},
"pg": {
"xact_commit": 17740,
"xact_rollback": 0,
"tup_inserted": 8794,
"blks_read": 0,
"blks_hit": 210015,
"wal_records": 62313,
"wal_bytes": 5534011
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4493,
"total_ms": 1006.2,
"mean_ms": 0.2239,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4493,
"total_ms": 420.3,
"mean_ms": 0.0935,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 8986,
"total_ms": 318.4,
"mean_ms": 0.0354,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4493,
"total_ms": 257.7,
"mean_ms": 0.0573,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 4493,
"total_ms": 86.6,
"mean_ms": 0.0193,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9033,
"total_ms": 12.3,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 9033,
"total_ms": 11.5,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 47,
"total_ms": 0.2,
"mean_ms": 0.0047,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0543,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1052,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0231,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.92,
"statements_per_ok": 15.04,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3971,
"ok": 3971,
"dropped": 30,
"ok_rps": 198.6,
"error_ratio": 0,
"ok_p50_ms": 14.164644,
"ok_p95_ms": 155.16024099999998,
"ok_p99_ms": 257.88778990000014,
"vus_max": 108,
"cpu": {
"api_total": {
"cpu_median": 103.2,
"cpu_max": 143.2,
"samples": 10
},
"api-1": {
"cpu_median": 103.2,
"cpu_max": 143.2,
"mem_last": "69.62MiB",
"samples": 10
},
"k6": {
"cpu_median": 22.7,
"cpu_max": 26.2,
"mem_last": "48.87MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.9,
"cpu_max": 17.1,
"mem_last": "12.14MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 45.1,
"cpu_max": 51.0,
"mem_last": "78.31MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 6.1,
"cpu_max": 6.8,
"mem_last": "3.668MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7976,
"pool_wait_ms_avg": 5.877,
"pool_waiting_max": 26,
"pool_usage_ms_avg": 8.018,
"api_cpu_s": 18.75,
"intervals": 10
},
"pg": {
"xact_commit": 23634,
"xact_rollback": 0,
"tup_inserted": 11790,
"blks_read": 0,
"blks_hit": 283077,
"wal_records": 83272,
"wal_bytes": 7352333
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5931,
"total_ms": 1357.9,
"mean_ms": 0.2289,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5931,
"total_ms": 652.6,
"mean_ms": 0.11,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5931,
"total_ms": 405.9,
"mean_ms": 0.0684,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11862,
"total_ms": 397.7,
"mean_ms": 0.0335,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5931,
"total_ms": 142.2,
"mean_ms": 0.024,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11908,
"total_ms": 14.3,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 11908,
"total_ms": 13.1,
"mean_ms": 0.0011,
"query": "BEGIN"
},
{
"calls": 46,
"total_ms": 0.2,
"mean_ms": 0.0049,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1403,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.041,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0227,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.95,
"statements_per_ok": 14.97,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4396,
"ok": 4396,
"dropped": 604,
"ok_rps": 219.8,
"error_ratio": 0,
"ok_p50_ms": 1792.2371604999998,
"ok_p95_ms": 2446.1017865,
"ok_p99_ms": 2502.0942151000004,
"vus_max": 773,
"cpu": {
"api_total": {
"cpu_median": 100.6,
"cpu_max": 142.2,
"samples": 10
},
"api-1": {
"cpu_median": 100.6,
"cpu_max": 142.2,
"mem_last": "63.25MiB",
"samples": 10
},
"k6": {
"cpu_median": 28.7,
"cpu_max": 43.8,
"mem_last": "227.6MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.4,
"cpu_max": 19.6,
"mem_last": "18.64MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 47.1,
"cpu_max": 68.6,
"mem_last": "82.79MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 5.8,
"cpu_max": 6.5,
"mem_last": "3.797MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8375,
"pool_wait_ms_avg": 446.111,
"pool_waiting_max": 188,
"pool_usage_ms_avg": 17.176,
"api_cpu_s": 20.01,
"intervals": 10
},
"pg": {
"xact_commit": 26468,
"xact_rollback": 0,
"tup_inserted": 13076,
"blks_read": 0,
"blks_hit": 324941,
"wal_records": 92938,
"wal_bytes": 8301132
},
"lb": {
"chkdown_delta": 0,
"qmax": 300,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6647,
"total_ms": 1537.9,
"mean_ms": 0.2314,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6647,
"total_ms": 732.0,
"mean_ms": 0.1101,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6647,
"total_ms": 474.8,
"mean_ms": 0.0714,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 13294,
"total_ms": 410.1,
"mean_ms": 0.0308,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6647,
"total_ms": 166.9,
"mean_ms": 0.0251,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13334,
"total_ms": 15.7,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 13334,
"total_ms": 13.1,
"mean_ms": 0.001,
"query": "BEGIN"
},
{
"calls": 40,
"total_ms": 0.2,
"mean_ms": 0.0044,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0594,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1177,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0272,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.02,
"statements_per_ok": 15.15,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 4201,
"ok": 4201,
"dropped": 1800,
"ok_rps": 210.1,
"error_ratio": 0,
"ok_p50_ms": 2729.610138,
"ok_p95_ms": 2994.333622,
"ok_p99_ms": 3070.33987,
"vus_max": 1139,
"cpu": {
"api_total": {
"cpu_median": 105.5,
"cpu_max": 150.1,
"samples": 10
},
"api-1": {
"cpu_median": 105.5,
"cpu_max": 150.1,
"mem_last": "65.55MiB",
"samples": 10
},
"k6": {
"cpu_median": 27.2,
"cpu_max": 44.6,
"mem_last": "366MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.4,
"cpu_max": 20.7,
"mem_last": "21.2MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 46.6,
"cpu_max": 49.5,
"mem_last": "87.3MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 5.9,
"cpu_max": 8.1,
"mem_last": "4.004MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8419,
"pool_wait_ms_avg": 444.989,
"pool_waiting_max": 187,
"pool_usage_ms_avg": 17.114,
"api_cpu_s": 20.05,
"intervals": 10
},
"pg": {
"xact_commit": 27032,
"xact_rollback": 0,
"tup_inserted": 13484,
"blks_read": 0,
"blks_hit": 345396,
"wal_records": 94710,
"wal_bytes": 8588339
},
"lb": {
"chkdown_delta": 0,
"qmax": 400,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 27,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6670,
"total_ms": 1537.3,
"mean_ms": 0.2305,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6670,
"total_ms": 751.4,
"mean_ms": 0.1127,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6670,
"total_ms": 476.1,
"mean_ms": 0.0714,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 13340,
"total_ms": 410.3,
"mean_ms": 0.0308,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6670,
"total_ms": 166.3,
"mean_ms": 0.0249,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13378,
"total_ms": 15.6,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 13378,
"total_ms": 13.0,
"mean_ms": 0.001,
"query": "BEGIN"
},
{
"calls": 38,
"total_ms": 0.2,
"mean_ms": 0.005,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.056,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0969,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0193,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.43,
"statements_per_ok": 15.91,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api2-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 3975,
"ok": 3975,
"dropped": 3026,
"ok_rps": 198.8,
"error_ratio": 0,
"ok_p50_ms": 2965.12569,
"ok_p95_ms": 3324.3593067,
"ok_p99_ms": 3363.43487784,
"vus_max": 1200,
"cpu": {
"api_total": {
"cpu_median": 115.7,
"cpu_max": 139.8,
"samples": 10
},
"api-1": {
"cpu_median": 115.7,
"cpu_max": 139.8,
"mem_last": "66.34MiB",
"samples": 10
},
"k6": {
"cpu_median": 26.5,
"cpu_max": 67.9,
"mem_last": "350MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.0,
"cpu_max": 19.2,
"mem_last": "22.2MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.4,
"cpu_max": 48.6,
"mem_last": "91.29MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.1,
"cpu_max": 2.2,
"mem_last": "3.969MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7975,
"pool_wait_ms_avg": 467.16,
"pool_waiting_max": 188,
"pool_usage_ms_avg": 18.209,
"api_cpu_s": 20.02,
"intervals": 10
},
"pg": {
"xact_commit": 25821,
"xact_rollback": 0,
"tup_inserted": 12830,
"blks_read": 0,
"blks_hit": 333920,
"wal_records": 90730,
"wal_bytes": 7895331
},
"lb": {
"chkdown_delta": 0,
"qmax": 411,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 23,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6448,
"total_ms": 1522.2,
"mean_ms": 0.2361,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6448,
"total_ms": 737.4,
"mean_ms": 0.1144,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 6448,
"total_ms": 476.4,
"mean_ms": 0.0739,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 12896,
"total_ms": 402.3,
"mean_ms": 0.0312,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6448,
"total_ms": 165.8,
"mean_ms": 0.0257,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 12935,
"total_ms": 15.7,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 12935,
"total_ms": 12.9,
"mean_ms": 0.001,
"query": "BEGIN"
},
{
"calls": 39,
"total_ms": 0.2,
"mean_ms": 0.0046,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1453,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0516,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0217,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.5,
"statements_per_ok": 16.25,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 2.0,
"memory_mib": 768,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2001,
"ok": 2001,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 7.80253,
"ok_p95_ms": 31.97001,
"ok_p99_ms": 59.668254,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 83.4,
"cpu_max": 143.0,
"samples": 9
},
"api-1": {
"cpu_median": 42.3,
"cpu_max": 72.0,
"mem_last": "55.59MiB",
"samples": 9
},
"api-2": {
"cpu_median": 42.1,
"cpu_max": 70.9,
"mem_last": "57.05MiB",
"samples": 9
},
"k6": {
"cpu_median": 12.1,
"cpu_max": 13.5,
"mem_last": "25.66MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 4.1,
"cpu_max": 13.7,
"mem_last": "12.02MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 21.0,
"cpu_max": 27.8,
"mem_last": "71.45MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 5.2,
"cpu_max": 6.2,
"mem_last": "3.828MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 4054,
"pool_wait_ms_avg": 0.001,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 2.909,
"api_cpu_s": 10.83,
"intervals": 20
},
"pg": {
"xact_commit": 12104,
"xact_rollback": 0,
"tup_inserted": 6019,
"blks_read": 60,
"blks_hit": 127475,
"wal_records": 42792,
"wal_bytes": 4214586
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3001,
"total_ms": 719.3,
"mean_ms": 0.2397,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3001,
"total_ms": 320.8,
"mean_ms": 0.1069,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 3001,
"total_ms": 191.1,
"mean_ms": 0.0637,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 6002,
"total_ms": 150.9,
"mean_ms": 0.0251,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3001,
"total_ms": 59.1,
"mean_ms": 0.0197,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6096,
"total_ms": 8.4,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 6096,
"total_ms": 7.7,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 94,
"total_ms": 0.6,
"mean_ms": 0.006,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1248,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0462,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0228,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.05,
"statements_per_ok": 15.14,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2993,
"ok": 2993,
"dropped": 8,
"ok_rps": 149.7,
"error_ratio": 0,
"ok_p50_ms": 10.136685,
"ok_p95_ms": 84.46405700000001,
"ok_p99_ms": 245.7217186799998,
"vus_max": 55,
"cpu": {
"api_total": {
"cpu_median": 121.3,
"cpu_max": 181.5,
"samples": 10
},
"api-1": {
"cpu_median": 57.6,
"cpu_max": 85.2,
"mem_last": "56.37MiB",
"samples": 10
},
"api-2": {
"cpu_median": 59.5,
"cpu_max": 100.1,
"mem_last": "57.36MiB",
"samples": 10
},
"k6": {
"cpu_median": 19.1,
"cpu_max": 24.7,
"mem_last": "30.44MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 6.6,
"cpu_max": 18.2,
"mem_last": "11.93MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 39.0,
"cpu_max": 55.2,
"mem_last": "99.55MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 5.6,
"cpu_max": 8.7,
"mem_last": "3.5MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6045,
"pool_wait_ms_avg": 0.528,
"pool_waiting_max": 1,
"pool_usage_ms_avg": 6.071,
"api_cpu_s": 16.58,
"intervals": 20
},
"pg": {
"xact_commit": 17910,
"xact_rollback": 0,
"tup_inserted": 8844,
"blks_read": 0,
"blks_hit": 187600,
"wal_records": 62703,
"wal_bytes": 5584015
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4486,
"total_ms": 1154.9,
"mean_ms": 0.2574,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4486,
"total_ms": 504.7,
"mean_ms": 0.1125,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 4486,
"total_ms": 309.1,
"mean_ms": 0.0689,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8972,
"total_ms": 227.6,
"mean_ms": 0.0254,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4486,
"total_ms": 98.5,
"mean_ms": 0.022,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9062,
"total_ms": 12.4,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 9062,
"total_ms": 11.8,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 90,
"total_ms": 0.5,
"mean_ms": 0.0056,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1001,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0395,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.024,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.98,
"statements_per_ok": 15.08,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3952,
"ok": 3952,
"dropped": 49,
"ok_rps": 197.6,
"error_ratio": 0,
"ok_p50_ms": 15.75066,
"ok_p95_ms": 307.54845785,
"ok_p99_ms": 356.5846853,
"vus_max": 132,
"cpu": {
"api_total": {
"cpu_median": 151.1,
"cpu_max": 206.9,
"samples": 9
},
"api-1": {
"cpu_median": 75.7,
"cpu_max": 101.0,
"mem_last": "56.96MiB",
"samples": 9
},
"api-2": {
"cpu_median": 75.4,
"cpu_max": 105.9,
"mem_last": "58.1MiB",
"samples": 9
},
"k6": {
"cpu_median": 26.8,
"cpu_max": 31.6,
"mem_last": "56.85MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 10.1,
"cpu_max": 18.5,
"mem_last": "12.45MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 52.6,
"cpu_max": 69.4,
"mem_last": "102.9MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 6.5,
"cpu_max": 9.6,
"mem_last": "3.734MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 7968,
"pool_wait_ms_avg": 8.965,
"pool_waiting_max": 16,
"pool_usage_ms_avg": 12.924,
"api_cpu_s": 22.83,
"intervals": 20
},
"pg": {
"xact_commit": 23831,
"xact_rollback": 0,
"tup_inserted": 11818,
"blks_read": 0,
"blks_hit": 257503,
"wal_records": 83528,
"wal_bytes": 7386866
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5906,
"total_ms": 1628.1,
"mean_ms": 0.2757,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5906,
"total_ms": 739.6,
"mean_ms": 0.1252,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5906,
"total_ms": 493.7,
"mean_ms": 0.0836,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11812,
"total_ms": 303.2,
"mean_ms": 0.0257,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5906,
"total_ms": 153.5,
"mean_ms": 0.026,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11900,
"total_ms": 16.5,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 11900,
"total_ms": 16.4,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 88,
"total_ms": 0.4,
"mean_ms": 0.0048,
"query": "SELECT $1"
},
{
"calls": 2,
"total_ms": 0.2,
"mean_ms": 0.0981,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1129,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0172,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.03,
"statements_per_ok": 15.01,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4848,
"ok": 4848,
"dropped": 152,
"ok_rps": 242.4,
"error_ratio": 0,
"ok_p50_ms": 61.385312,
"ok_p95_ms": 630.4609458499998,
"ok_p99_ms": 819.2782447999998,
"vus_max": 287,
"cpu": {
"api_total": {
"cpu_median": 173.5,
"cpu_max": 197.5,
"samples": 10
},
"api-1": {
"cpu_median": 87.2,
"cpu_max": 98.4,
"mem_last": "57.06MiB",
"samples": 10
},
"api-2": {
"cpu_median": 86.4,
"cpu_max": 99.1,
"mem_last": "58.86MiB",
"samples": 10
},
"k6": {
"cpu_median": 32.4,
"cpu_max": 38.9,
"mem_last": "90.46MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 11.7,
"cpu_max": 25.3,
"mem_last": "13.41MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 53.4,
"cpu_max": 67.7,
"mem_last": "107.9MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 7.4,
"cpu_max": 8.7,
"mem_last": "4.164MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 9352,
"pool_wait_ms_avg": 38.644,
"pool_waiting_max": 53,
"pool_usage_ms_avg": 16.382,
"api_cpu_s": 27.06,
"intervals": 19
},
"pg": {
"xact_commit": 28717,
"xact_rollback": 0,
"tup_inserted": 14204,
"blks_read": 0,
"blks_hit": 318196,
"wal_records": 100699,
"wal_bytes": 9071845
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 7,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7248,
"total_ms": 2091.8,
"mean_ms": 0.2886,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7248,
"total_ms": 938.7,
"mean_ms": 0.1295,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 7248,
"total_ms": 651.3,
"mean_ms": 0.0899,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 14496,
"total_ms": 348.4,
"mean_ms": 0.024,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7248,
"total_ms": 197.6,
"mean_ms": 0.0273,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14580,
"total_ms": 20.5,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 14580,
"total_ms": 20.0,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 84,
"total_ms": 0.4,
"mean_ms": 0.0047,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1207,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0485,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0186,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.92,
"statements_per_ok": 15.0,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 5629,
"ok": 5629,
"dropped": 372,
"ok_rps": 281.4,
"error_ratio": 0,
"ok_p50_ms": 785.742226,
"ok_p95_ms": 1390.0188632,
"ok_p99_ms": 1561.0473299599998,
"vus_max": 654,
"cpu": {
"api_total": {
"cpu_median": 197.3,
"cpu_max": 199.8,
"samples": 10
},
"api-1": {
"cpu_median": 97.8,
"cpu_max": 100.8,
"mem_last": "74.71MiB",
"samples": 10
},
"api-2": {
"cpu_median": 98.8,
"cpu_max": 100.6,
"mem_last": "74.56MiB",
"samples": 10
},
"k6": {
"cpu_median": 36.9,
"cpu_max": 54.5,
"mem_last": "187.3MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 13.3,
"cpu_max": 28.0,
"mem_last": "17.03MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 62.9,
"cpu_max": 83.1,
"mem_last": "113.2MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 7.8,
"cpu_max": 9.4,
"mem_last": "3.785MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 11313,
"pool_wait_ms_avg": 306.747,
"pool_waiting_max": 178,
"pool_usage_ms_avg": 25.547,
"api_cpu_s": 33.38,
"intervals": 20
},
"pg": {
"xact_commit": 33657,
"xact_rollback": 0,
"tup_inserted": 16620,
"blks_read": 0,
"blks_hit": 373365,
"wal_records": 116988,
"wal_bytes": 10236889
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 4,
"redispatch_delta": 0
},
"statements": [
{
"calls": 8360,
"total_ms": 2433.7,
"mean_ms": 0.2911,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 8360,
"total_ms": 1133.0,
"mean_ms": 0.1355,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 8360,
"total_ms": 877.6,
"mean_ms": 0.105,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 16720,
"total_ms": 390.1,
"mean_ms": 0.0233,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 8360,
"total_ms": 233.0,
"mean_ms": 0.0279,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 16792,
"total_ms": 22.1,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 16792,
"total_ms": 22.0,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 72,
"total_ms": 0.3,
"mean_ms": 0.0048,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0616,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1172,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0242,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.98,
"statements_per_ok": 14.89,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 5423,
"ok": 5423,
"dropped": 1577,
"ok_rps": 271.1,
"error_ratio": 0,
"ok_p50_ms": 1980.052642,
"ok_p95_ms": 2614.7839925999997,
"ok_p99_ms": 2785.22204186,
"vus_max": 1061,
"cpu": {
"api_total": {
"cpu_median": 196.2,
"cpu_max": 206.4,
"samples": 10
},
"api-1": {
"cpu_median": 98.5,
"cpu_max": 105.5,
"mem_last": "76.6MiB",
"samples": 10
},
"api-2": {
"cpu_median": 99.4,
"cpu_max": 102.5,
"mem_last": "75.66MiB",
"samples": 10
},
"k6": {
"cpu_median": 44.3,
"cpu_max": 77.3,
"mem_last": "375.3MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 15.6,
"cpu_max": 31.1,
"mem_last": "20.72MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 65.1,
"cpu_max": 86.2,
"mem_last": "118.5MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 3.0,
"cpu_max": 3.9,
"mem_last": "3.746MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 10732,
"pool_wait_ms_avg": 692.084,
"pool_waiting_max": 190,
"pool_usage_ms_avg": 27.437,
"api_cpu_s": 33.31,
"intervals": 20
},
"pg": {
"xact_commit": 33236,
"xact_rollback": 0,
"tup_inserted": 16476,
"blks_read": 0,
"blks_hit": 377844,
"wal_records": 118162,
"wal_bytes": 10573711
},
"lb": {
"chkdown_delta": 0,
"qmax": 200,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 6,
"redispatch_delta": 0
},
"statements": [
{
"calls": 8461,
"total_ms": 3103.9,
"mean_ms": 0.3668,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8461,
"total_ms": 2680.9,
"mean_ms": 0.3169,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 8461,
"total_ms": 1199.9,
"mean_ms": 0.1418,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 16922,
"total_ms": 421.1,
"mean_ms": 0.0249,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 8461,
"total_ms": 259.9,
"mean_ms": 0.0307,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 16991,
"total_ms": 27.6,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 16991,
"total_ms": 22.4,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 69,
"total_ms": 0.4,
"mean_ms": 0.0056,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.131,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0397,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.022,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.13,
"statements_per_ok": 15.64,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2000,
"ok": 2000,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 8.537046,
"ok_p95_ms": 28.476515,
"ok_p99_ms": 43.90845024,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 86.2,
"cpu_max": 131.7,
"samples": 10
},
"api-1": {
"cpu_median": 43.0,
"cpu_max": 65.2,
"mem_last": "55.7MiB",
"samples": 10
},
"api-2": {
"cpu_median": 43.2,
"cpu_max": 66.4,
"mem_last": "55.7MiB",
"samples": 10
},
"k6": {
"cpu_median": 11.7,
"cpu_max": 14.5,
"mem_last": "25.63MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 4.3,
"cpu_max": 13.0,
"mem_last": "11.78MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 19.9,
"cpu_max": 33.2,
"mem_last": "66.48MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.5,
"cpu_max": 4.9,
"mem_last": "3.688MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 4057,
"pool_wait_ms_avg": 0.0,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 2.955,
"api_cpu_s": 10.43,
"intervals": 20
},
"pg": {
"xact_commit": 12103,
"xact_rollback": 0,
"tup_inserted": 6031,
"blks_read": 83,
"blks_hit": 131123,
"wal_records": 42924,
"wal_bytes": 4205568
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3001,
"total_ms": 728.9,
"mean_ms": 0.2429,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3001,
"total_ms": 312.2,
"mean_ms": 0.104,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 3001,
"total_ms": 181.1,
"mean_ms": 0.0603,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 6002,
"total_ms": 149.6,
"mean_ms": 0.0249,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3001,
"total_ms": 55.7,
"mean_ms": 0.0186,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6096,
"total_ms": 8.7,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 6096,
"total_ms": 8.3,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 94,
"total_ms": 0.5,
"mean_ms": 0.0053,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.2,
"mean_ms": 0.1093,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.1597,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0241,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.05,
"statements_per_ok": 15.15,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 3001,
"ok": 3001,
"dropped": 0,
"ok_rps": 150.1,
"error_ratio": 0,
"ok_p50_ms": 9.827698,
"ok_p95_ms": 61.054989,
"ok_p99_ms": 102.130968,
"vus_max": 59,
"cpu": {
"api_total": {
"cpu_median": 93.4,
"cpu_max": 173.7,
"samples": 9
},
"api-1": {
"cpu_median": 40.2,
"cpu_max": 90.9,
"mem_last": "55.75MiB",
"samples": 9
},
"api-2": {
"cpu_median": 44.7,
"cpu_max": 86.3,
"mem_last": "55.78MiB",
"samples": 9
},
"k6": {
"cpu_median": 18.7,
"cpu_max": 20.5,
"mem_last": "31.87MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 6.2,
"cpu_max": 15.8,
"mem_last": "11.93MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 29.6,
"cpu_max": 37.1,
"mem_last": "99.54MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 5.7,
"cpu_max": 6.3,
"mem_last": "3.758MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 6057,
"pool_wait_ms_avg": 0.002,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 4.759,
"api_cpu_s": 16.74,
"intervals": 20
},
"pg": {
"xact_commit": 17935,
"xact_rollback": 0,
"tup_inserted": 8902,
"blks_read": 0,
"blks_hit": 197347,
"wal_records": 62727,
"wal_bytes": 5537639
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4483,
"total_ms": 1177.2,
"mean_ms": 0.2626,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4483,
"total_ms": 483.7,
"mean_ms": 0.1079,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 4483,
"total_ms": 284.1,
"mean_ms": 0.0634,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8966,
"total_ms": 226.7,
"mean_ms": 0.0253,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4483,
"total_ms": 90.5,
"mean_ms": 0.0202,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9057,
"total_ms": 13.3,
"mean_ms": 0.0015,
"query": "BEGIN"
},
{
"calls": 9057,
"total_ms": 13.1,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 91,
"total_ms": 0.5,
"mean_ms": 0.0051,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1075,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0527,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0267,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.98,
"statements_per_ok": 15.03,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3970,
"ok": 3970,
"dropped": 31,
"ok_rps": 198.5,
"error_ratio": 0,
"ok_p50_ms": 11.526496,
"ok_p95_ms": 198.70631769999994,
"ok_p99_ms": 318.32055527,
"vus_max": 114,
"cpu": {
"api_total": {
"cpu_median": 138.3,
"cpu_max": 195.3,
"samples": 10
},
"api-1": {
"cpu_median": 75.2,
"cpu_max": 94.8,
"mem_last": "67.4MiB",
"samples": 10
},
"api-2": {
"cpu_median": 61.5,
"cpu_max": 101.7,
"mem_last": "65.75MiB",
"samples": 10
},
"k6": {
"cpu_median": 26.9,
"cpu_max": 30.8,
"mem_last": "52.26MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 9.0,
"cpu_max": 17.3,
"mem_last": "12.24MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 42.9,
"cpu_max": 60.1,
"mem_last": "103.1MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 4.9,
"cpu_max": 9.7,
"mem_last": "3.859MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 7956,
"pool_wait_ms_avg": 5.568,
"pool_waiting_max": 22,
"pool_usage_ms_avg": 10.138,
"api_cpu_s": 22.42,
"intervals": 20
},
"pg": {
"xact_commit": 23820,
"xact_rollback": 0,
"tup_inserted": 11802,
"blks_read": 0,
"blks_hit": 265814,
"wal_records": 83330,
"wal_bytes": 7405507
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5922,
"total_ms": 1576.8,
"mean_ms": 0.2663,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5922,
"total_ms": 724.1,
"mean_ms": 0.1223,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5922,
"total_ms": 490.7,
"mean_ms": 0.0829,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11844,
"total_ms": 294.4,
"mean_ms": 0.0249,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5922,
"total_ms": 149.4,
"mean_ms": 0.0252,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11932,
"total_ms": 16.8,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 11932,
"total_ms": 16.5,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 88,
"total_ms": 0.4,
"mean_ms": 0.0047,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0591,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0939,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0244,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.0,
"statements_per_ok": 14.98,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4922,
"ok": 4922,
"dropped": 79,
"ok_rps": 246.1,
"error_ratio": 0,
"ok_p50_ms": 20.5787295,
"ok_p95_ms": 382.6286063499999,
"ok_p99_ms": 488.18310679,
"vus_max": 183,
"cpu": {
"api_total": {
"cpu_median": 163.3,
"cpu_max": 198.0,
"samples": 9
},
"api-1": {
"cpu_median": 81.7,
"cpu_max": 98.6,
"mem_last": "62.39MiB",
"samples": 9
},
"api-2": {
"cpu_median": 77.4,
"cpu_max": 99.4,
"mem_last": "57.39MiB",
"samples": 9
},
"k6": {
"cpu_median": 32.1,
"cpu_max": 37.4,
"mem_last": "74.4MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 11.2,
"cpu_max": 23.4,
"mem_last": "12.82MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 54.7,
"cpu_max": 60.7,
"mem_last": "107.8MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 7.4,
"cpu_max": 9.1,
"mem_last": "3.852MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 9942,
"pool_wait_ms_avg": 25.254,
"pool_waiting_max": 46,
"pool_usage_ms_avg": 13.996,
"api_cpu_s": 27.72,
"intervals": 20
},
"pg": {
"xact_commit": 29475,
"xact_rollback": 0,
"tup_inserted": 14646,
"blks_read": 0,
"blks_hit": 357956,
"wal_records": 103601,
"wal_bytes": 9290226
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7364,
"total_ms": 2065.4,
"mean_ms": 0.2805,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7364,
"total_ms": 938.2,
"mean_ms": 0.1274,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 7364,
"total_ms": 677.4,
"mean_ms": 0.092,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 14728,
"total_ms": 347.0,
"mean_ms": 0.0236,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7364,
"total_ms": 193.4,
"mean_ms": 0.0263,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14816,
"total_ms": 20.7,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 14816,
"total_ms": 20.0,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 88,
"total_ms": 0.4,
"mean_ms": 0.0045,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.2,
"mean_ms": 0.1066,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1147,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0232,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.99,
"statements_per_ok": 15.02,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 5813,
"ok": 5813,
"dropped": 188,
"ok_rps": 290.6,
"error_ratio": 0,
"ok_p50_ms": 344.621426,
"ok_p95_ms": 703.3058816,
"ok_p99_ms": 966.31137016,
"vus_max": 348,
"cpu": {
"api_total": {
"cpu_median": 196.3,
"cpu_max": 203.3,
"samples": 9
},
"api-1": {
"cpu_median": 98.9,
"cpu_max": 102.5,
"mem_last": "61.89MiB",
"samples": 9
},
"api-2": {
"cpu_median": 98.4,
"cpu_max": 101.0,
"mem_last": "67.79MiB",
"samples": 9
},
"k6": {
"cpu_median": 38.4,
"cpu_max": 48.5,
"mem_last": "134.9MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 14.9,
"cpu_max": 27.3,
"mem_last": "14.32MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 80.7,
"cpu_max": 86.0,
"mem_last": "113.1MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 3.6,
"cpu_max": 8.1,
"mem_last": "3.805MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 11720,
"pool_wait_ms_avg": 129.1,
"pool_waiting_max": 86,
"pool_usage_ms_avg": 23.613,
"api_cpu_s": 33.66,
"intervals": 20
},
"pg": {
"xact_commit": 34419,
"xact_rollback": 0,
"tup_inserted": 17058,
"blks_read": 0,
"blks_hit": 426282,
"wal_records": 120053,
"wal_bytes": 10557801
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 8,
"redispatch_delta": 0
},
"statements": [
{
"calls": 8698,
"total_ms": 2557.9,
"mean_ms": 0.2941,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 8698,
"total_ms": 1160.0,
"mean_ms": 0.1334,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 8698,
"total_ms": 896.7,
"mean_ms": 0.1031,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 17396,
"total_ms": 392.2,
"mean_ms": 0.0225,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 8698,
"total_ms": 240.7,
"mean_ms": 0.0277,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 17479,
"total_ms": 26.7,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 17479,
"total_ms": 22.3,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 83,
"total_ms": 0.4,
"mean_ms": 0.0045,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1141,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.053,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0225,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.92,
"statements_per_ok": 15.01,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 5813,
"ok": 5813,
"dropped": 1187,
"ok_rps": 290.6,
"error_ratio": 0,
"ok_p50_ms": 1844.997341,
"ok_p95_ms": 2348.4341228000003,
"ok_p99_ms": 2418.85193552,
"vus_max": 1052,
"cpu": {
"api_total": {
"cpu_median": 197.9,
"cpu_max": 203.5,
"samples": 10
},
"api-1": {
"cpu_median": 98.4,
"cpu_max": 102.0,
"mem_last": "62.18MiB",
"samples": 10
},
"api-2": {
"cpu_median": 99.4,
"cpu_max": 103.9,
"mem_last": "62.23MiB",
"samples": 10
},
"k6": {
"cpu_median": 42.0,
"cpu_max": 70.7,
"mem_last": "354.4MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 14.2,
"cpu_max": 28.8,
"mem_last": "20.29MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 65.6,
"cpu_max": 83.6,
"mem_last": "118.5MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 3.5,
"cpu_max": 4.1,
"mem_last": "4.145MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 11355,
"pool_wait_ms_avg": 610.592,
"pool_waiting_max": 188,
"pool_usage_ms_avg": 25.917,
"api_cpu_s": 34.3,
"intervals": 20
},
"pg": {
"xact_commit": 35767,
"xact_rollback": 0,
"tup_inserted": 17686,
"blks_read": 0,
"blks_hit": 449062,
"wal_records": 124812,
"wal_bytes": 11093891
},
"lb": {
"chkdown_delta": 0,
"qmax": 200,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 18,
"redispatch_delta": 0
},
"statements": [
{
"calls": 8832,
"total_ms": 2688.5,
"mean_ms": 0.3044,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 8832,
"total_ms": 1356.3,
"mean_ms": 0.1536,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8832,
"total_ms": 1209.3,
"mean_ms": 0.1369,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 17664,
"total_ms": 402.8,
"mean_ms": 0.0228,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 8832,
"total_ms": 247.1,
"mean_ms": 0.028,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 17734,
"total_ms": 25.1,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 17734,
"total_ms": 22.1,
"mean_ms": 0.0012,
"query": "BEGIN"
},
{
"calls": 70,
"total_ms": 0.4,
"mean_ms": 0.0054,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.2,
"mean_ms": 0.0891,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0965,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0226,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.15,
"statements_per_ok": 15.23,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2001,
"ok": 2001,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 6.980185,
"ok_p95_ms": 25.244786,
"ok_p99_ms": 42.937627,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 77.5,
"cpu_max": 132.4,
"samples": 9
},
"api-1": {
"cpu_median": 41.2,
"cpu_max": 67.1,
"mem_last": "56.93MiB",
"samples": 9
},
"api-2": {
"cpu_median": 36.3,
"cpu_max": 65.3,
"mem_last": "55.71MiB",
"samples": 9
},
"k6": {
"cpu_median": 11.4,
"cpu_max": 12.6,
"mem_last": "25.39MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 4.2,
"cpu_max": 11.9,
"mem_last": "11.8MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 18.7,
"cpu_max": 26.1,
"mem_last": "66.31MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 4.6,
"cpu_max": 5.9,
"mem_last": "3.422MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 4057,
"pool_wait_ms_avg": 0.0,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 2.406,
"api_cpu_s": 10.02,
"intervals": 20
},
"pg": {
"xact_commit": 12168,
"xact_rollback": 0,
"tup_inserted": 6075,
"blks_read": 60,
"blks_hit": 127589,
"wal_records": 43312,
"wal_bytes": 4247596
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3002,
"total_ms": 678.1,
"mean_ms": 0.2259,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3002,
"total_ms": 293.3,
"mean_ms": 0.0977,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 3002,
"total_ms": 178.5,
"mean_ms": 0.0595,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 6004,
"total_ms": 146.1,
"mean_ms": 0.0243,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3002,
"total_ms": 49.5,
"mean_ms": 0.0165,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6098,
"total_ms": 7.6,
"mean_ms": 0.0012,
"query": "BEGIN"
},
{
"calls": 6098,
"total_ms": 7.1,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 94,
"total_ms": 0.5,
"mean_ms": 0.0049,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1353,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0493,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0333,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.08,
"statements_per_ok": 15.15,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2984,
"ok": 2984,
"dropped": 17,
"ok_rps": 149.2,
"error_ratio": 0,
"ok_p50_ms": 8.575164000000001,
"ok_p95_ms": 96.4983846999999,
"ok_p99_ms": 263.7259766300002,
"vus_max": 65,
"cpu": {
"api_total": {
"cpu_median": 109.1,
"cpu_max": 200.0,
"samples": 10
},
"api-1": {
"cpu_median": 53.7,
"cpu_max": 99.4,
"mem_last": "68.94MiB",
"samples": 10
},
"api-2": {
"cpu_median": 55.1,
"cpu_max": 100.6,
"mem_last": "68.79MiB",
"samples": 10
},
"k6": {
"cpu_median": 16.6,
"cpu_max": 27.2,
"mem_last": "29.31MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 6.2,
"cpu_max": 21.3,
"mem_last": "11.93MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 36.9,
"cpu_max": 51.7,
"mem_last": "99.55MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.3,
"cpu_max": 5.6,
"mem_last": "3.906MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6012,
"pool_wait_ms_avg": 0.335,
"pool_waiting_max": 2,
"pool_usage_ms_avg": 4.668,
"api_cpu_s": 15.87,
"intervals": 20
},
"pg": {
"xact_commit": 17944,
"xact_rollback": 0,
"tup_inserted": 8864,
"blks_read": 0,
"blks_hit": 189196,
"wal_records": 62935,
"wal_bytes": 5577453
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4475,
"total_ms": 1129.1,
"mean_ms": 0.2523,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4475,
"total_ms": 491.4,
"mean_ms": 0.1098,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 4475,
"total_ms": 300.8,
"mean_ms": 0.0672,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8950,
"total_ms": 241.3,
"mean_ms": 0.027,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4475,
"total_ms": 91.6,
"mean_ms": 0.0205,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9041,
"total_ms": 12.2,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 9041,
"total_ms": 11.6,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 91,
"total_ms": 0.5,
"mean_ms": 0.0058,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1188,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0431,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0198,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.01,
"statements_per_ok": 15.09,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3970,
"ok": 3970,
"dropped": 30,
"ok_rps": 198.5,
"error_ratio": 0,
"ok_p50_ms": 11.589998000000001,
"ok_p95_ms": 162.31980174999998,
"ok_p99_ms": 253.8206895199997,
"vus_max": 85,
"cpu": {
"api_total": {
"cpu_median": 110.8,
"cpu_max": 201.3,
"samples": 10
},
"api-1": {
"cpu_median": 57.7,
"cpu_max": 99.1,
"mem_last": "58.59MiB",
"samples": 10
},
"api-2": {
"cpu_median": 55.4,
"cpu_max": 102.5,
"mem_last": "61.45MiB",
"samples": 10
},
"k6": {
"cpu_median": 24.9,
"cpu_max": 27.7,
"mem_last": "41.13MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 8.8,
"cpu_max": 18.1,
"mem_last": "12.11MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.9,
"cpu_max": 48.7,
"mem_last": "103.2MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.6,
"cpu_max": 2.6,
"mem_last": "3.695MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8007,
"pool_wait_ms_avg": 1.351,
"pool_waiting_max": 6,
"pool_usage_ms_avg": 8.999,
"api_cpu_s": 21.45,
"intervals": 20
},
"pg": {
"xact_commit": 23812,
"xact_rollback": 0,
"tup_inserted": 11778,
"blks_read": 0,
"blks_hit": 246851,
"wal_records": 83290,
"wal_bytes": 7393664
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5956,
"total_ms": 1530.5,
"mean_ms": 0.257,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5956,
"total_ms": 708.6,
"mean_ms": 0.119,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5956,
"total_ms": 438.8,
"mean_ms": 0.0737,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11912,
"total_ms": 297.3,
"mean_ms": 0.025,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5956,
"total_ms": 146.3,
"mean_ms": 0.0246,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 12003,
"total_ms": 16.6,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 12003,
"total_ms": 15.8,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 91,
"total_ms": 0.4,
"mean_ms": 0.0044,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.1582,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0437,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0185,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.0,
"statements_per_ok": 15.07,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4901,
"ok": 4901,
"dropped": 100,
"ok_rps": 245.1,
"error_ratio": 0,
"ok_p50_ms": 38.657426,
"ok_p95_ms": 447.950927,
"ok_p99_ms": 561.220036,
"vus_max": 208,
"cpu": {
"api_total": {
"cpu_median": 153.7,
"cpu_max": 192.3,
"samples": 9
},
"api-1": {
"cpu_median": 73.0,
"cpu_max": 96.8,
"mem_last": "58.27MiB",
"samples": 9
},
"api-2": {
"cpu_median": 80.7,
"cpu_max": 95.8,
"mem_last": "57.45MiB",
"samples": 9
},
"k6": {
"cpu_median": 33.5,
"cpu_max": 35.9,
"mem_last": "78.45MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 11.0,
"cpu_max": 24.2,
"mem_last": "13.14MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 56.5,
"cpu_max": 67.2,
"mem_last": "108.4MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 3.3,
"cpu_max": 7.1,
"mem_last": "3.688MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 9826,
"pool_wait_ms_avg": 32.021,
"pool_waiting_max": 53,
"pool_usage_ms_avg": 15.9,
"api_cpu_s": 27.58,
"intervals": 20
},
"pg": {
"xact_commit": 29189,
"xact_rollback": 0,
"tup_inserted": 14470,
"blks_read": 0,
"blks_hit": 324088,
"wal_records": 102356,
"wal_bytes": 9182750
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7337,
"total_ms": 2044.5,
"mean_ms": 0.2787,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7337,
"total_ms": 937.2,
"mean_ms": 0.1277,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 7337,
"total_ms": 649.7,
"mean_ms": 0.0886,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 14674,
"total_ms": 350.3,
"mean_ms": 0.0239,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7337,
"total_ms": 198.1,
"mean_ms": 0.027,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14761,
"total_ms": 20.4,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 14761,
"total_ms": 19.5,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 87,
"total_ms": 0.4,
"mean_ms": 0.0045,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0736,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0988,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0358,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.96,
"statements_per_ok": 15.02,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 5718,
"ok": 5718,
"dropped": 283,
"ok_rps": 285.9,
"error_ratio": 0,
"ok_p50_ms": 641.7616855,
"ok_p95_ms": 1123.39635615,
"ok_p99_ms": 1214.89919999,
"vus_max": 451,
"cpu": {
"api_total": {
"cpu_median": 196.1,
"cpu_max": 201.6,
"samples": 9
},
"api-1": {
"cpu_median": 98.0,
"cpu_max": 102.5,
"mem_last": "65.95MiB",
"samples": 9
},
"api-2": {
"cpu_median": 98.4,
"cpu_max": 102.3,
"mem_last": "72.92MiB",
"samples": 9
},
"k6": {
"cpu_median": 35.8,
"cpu_max": 50.5,
"mem_last": "144.2MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 13.6,
"cpu_max": 22.9,
"mem_last": "21.56MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 66.5,
"cpu_max": 84.2,
"mem_last": "113.1MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 3.5,
"cpu_max": 3.6,
"mem_last": "3.637MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 11427,
"pool_wait_ms_avg": 252.69,
"pool_waiting_max": 165,
"pool_usage_ms_avg": 25.068,
"api_cpu_s": 32.95,
"intervals": 20
},
"pg": {
"xact_commit": 33957,
"xact_rollback": 0,
"tup_inserted": 16802,
"blks_read": 0,
"blks_hit": 377537,
"wal_records": 118622,
"wal_bytes": 10415770
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 7,
"redispatch_delta": 0
},
"statements": [
{
"calls": 8583,
"total_ms": 2462.6,
"mean_ms": 0.2869,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 8583,
"total_ms": 1167.7,
"mean_ms": 0.136,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8583,
"total_ms": 1152.6,
"mean_ms": 0.1343,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 17166,
"total_ms": 393.5,
"mean_ms": 0.0229,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 8583,
"total_ms": 241.1,
"mean_ms": 0.0281,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 17243,
"total_ms": 25.5,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 17243,
"total_ms": 22.2,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 77,
"total_ms": 0.5,
"mean_ms": 0.0065,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0996,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0376,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0206,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.94,
"statements_per_ok": 15.05,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,186 @@
{
"config": "n2-api1-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 5779,
"ok": 5779,
"dropped": 1221,
"ok_rps": 288.9,
"error_ratio": 0,
"ok_p50_ms": 1792.498706,
"ok_p95_ms": 2449.1132878999997,
"ok_p99_ms": 2635.6998854000003,
"vus_max": 1081,
"cpu": {
"api_total": {
"cpu_median": 201.3,
"cpu_max": 205.5,
"samples": 10
},
"api-1": {
"cpu_median": 99.4,
"cpu_max": 104.3,
"mem_last": "63.43MiB",
"samples": 10
},
"api-2": {
"cpu_median": 99.7,
"cpu_max": 103.6,
"mem_last": "63.29MiB",
"samples": 10
},
"k6": {
"cpu_median": 42.6,
"cpu_max": 83.7,
"mem_last": "337.2MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 13.8,
"cpu_max": 28.6,
"mem_last": "20.67MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 70.2,
"cpu_max": 90.2,
"mem_last": "118.6MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 8.0,
"cpu_max": 8.6,
"mem_last": "3.691MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 11457,
"pool_wait_ms_avg": 637.12,
"pool_waiting_max": 189,
"pool_usage_ms_avg": 25.678,
"api_cpu_s": 34.12,
"intervals": 20
},
"pg": {
"xact_commit": 35573,
"xact_rollback": 0,
"tup_inserted": 17664,
"blks_read": 0,
"blks_hit": 404301,
"wal_records": 125357,
"wal_bytes": 11082104
},
"lb": {
"chkdown_delta": 0,
"qmax": 200,
"up_after": 2,
"econ_delta": 0,
"eresp_delta": 24,
"redispatch_delta": 0
},
"statements": [
{
"calls": 8792,
"total_ms": 2660.1,
"mean_ms": 0.3026,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 8792,
"total_ms": 2546.3,
"mean_ms": 0.2896,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8792,
"total_ms": 1222.9,
"mean_ms": 0.1391,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 17584,
"total_ms": 413.5,
"mean_ms": 0.0235,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 8792,
"total_ms": 252.4,
"mean_ms": 0.0287,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 17654,
"total_ms": 27.1,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 17654,
"total_ms": 22.4,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 70,
"total_ms": 0.3,
"mean_ms": 0.0046,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0548,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.105,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0207,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.16,
"statements_per_ok": 15.25,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2001,
"ok": 2001,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 11.298931,
"ok_p95_ms": 77.948327,
"ok_p99_ms": 125.443179,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 110.8,
"cpu_max": 168.4,
"samples": 9
},
"api-1": {
"cpu_median": 28.6,
"cpu_max": 49.6,
"mem_last": "64.99MiB",
"samples": 9
},
"api-2": {
"cpu_median": 28.0,
"cpu_max": 51.4,
"mem_last": "66.88MiB",
"samples": 9
},
"api-3": {
"cpu_median": 24.3,
"cpu_max": 53.6,
"mem_last": "67.03MiB",
"samples": 9
},
"api-4": {
"cpu_median": 25.6,
"cpu_max": 49.8,
"mem_last": "54.93MiB",
"samples": 9
},
"k6": {
"cpu_median": 12.4,
"cpu_max": 16.5,
"mem_last": "25.34MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 4.6,
"cpu_max": 11.9,
"mem_last": "11.91MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 21.0,
"cpu_max": 39.9,
"mem_last": "88.5MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 1.5,
"cpu_max": 2.0,
"mem_last": "3.926MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 4090,
"pool_wait_ms_avg": 0.0,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 5.462,
"api_cpu_s": 11.39,
"intervals": 40
},
"pg": {
"xact_commit": 12361,
"xact_rollback": 0,
"tup_inserted": 6096,
"blks_read": 83,
"blks_hit": 135067,
"wal_records": 43319,
"wal_bytes": 4239661
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3002,
"total_ms": 751.8,
"mean_ms": 0.2504,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3002,
"total_ms": 340.0,
"mean_ms": 0.1132,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 3002,
"total_ms": 190.1,
"mean_ms": 0.0633,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 6004,
"total_ms": 154.2,
"mean_ms": 0.0257,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3002,
"total_ms": 73.2,
"mean_ms": 0.0244,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6186,
"total_ms": 8.4,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 6186,
"total_ms": 8.3,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 182,
"total_ms": 1.0,
"mean_ms": 0.0055,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.1591,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0675,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0271,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.18,
"statements_per_ok": 15.28,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2945,
"ok": 2945,
"dropped": 56,
"ok_rps": 147.2,
"error_ratio": 0,
"ok_p50_ms": 22.235219,
"ok_p95_ms": 460.2908489999996,
"ok_p99_ms": 789.53975552,
"vus_max": 103,
"cpu": {
"api_total": {
"cpu_median": 148.2,
"cpu_max": 213.5,
"samples": 10
},
"api-1": {
"cpu_median": 35.9,
"cpu_max": 52.9,
"mem_last": "55.95MiB",
"samples": 10
},
"api-2": {
"cpu_median": 39.0,
"cpu_max": 54.2,
"mem_last": "57.27MiB",
"samples": 10
},
"api-3": {
"cpu_median": 41.1,
"cpu_max": 53.9,
"mem_last": "57.51MiB",
"samples": 10
},
"api-4": {
"cpu_median": 37.4,
"cpu_max": 52.6,
"mem_last": "55.35MiB",
"samples": 10
},
"k6": {
"cpu_median": 19.9,
"cpu_max": 41.2,
"mem_last": "47.52MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.4,
"cpu_max": 18.4,
"mem_last": "12.6MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.3,
"cpu_max": 108.2,
"mem_last": "147.7MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 6.0,
"cpu_max": 13.6,
"mem_last": "3.75MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6009,
"pool_wait_ms_avg": 5.766,
"pool_waiting_max": 4,
"pool_usage_ms_avg": 21.43,
"api_cpu_s": 18.91,
"intervals": 40
},
"pg": {
"xact_commit": 17798,
"xact_rollback": 0,
"tup_inserted": 8702,
"blks_read": 0,
"blks_hit": 203829,
"wal_records": 61613,
"wal_bytes": 5482545
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4433,
"total_ms": 1476.1,
"mean_ms": 0.333,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4433,
"total_ms": 604.8,
"mean_ms": 0.1364,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 4433,
"total_ms": 408.9,
"mean_ms": 0.0922,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8866,
"total_ms": 249.3,
"mean_ms": 0.0281,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4433,
"total_ms": 108.7,
"mean_ms": 0.0245,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9030,
"total_ms": 19.3,
"mean_ms": 0.0021,
"query": "COMMIT"
},
{
"calls": 9030,
"total_ms": 13.1,
"mean_ms": 0.0015,
"query": "BEGIN"
},
{
"calls": 164,
"total_ms": 0.9,
"mean_ms": 0.0057,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1088,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0443,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0246,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.04,
"statements_per_ok": 15.22,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3872,
"ok": 3872,
"dropped": 129,
"ok_rps": 193.6,
"error_ratio": 0,
"ok_p50_ms": 180.1144295,
"ok_p95_ms": 751.6838958,
"ok_p99_ms": 982.1881199099998,
"vus_max": 205,
"cpu": {
"api_total": {
"cpu_median": 190.4,
"cpu_max": 195.5,
"samples": 9
},
"api-1": {
"cpu_median": 47.4,
"cpu_max": 51.7,
"mem_last": "68.65MiB",
"samples": 9
},
"api-2": {
"cpu_median": 46.7,
"cpu_max": 49.5,
"mem_last": "69.83MiB",
"samples": 9
},
"api-3": {
"cpu_median": 48.0,
"cpu_max": 50.5,
"mem_last": "68.59MiB",
"samples": 9
},
"api-4": {
"cpu_median": 46.1,
"cpu_max": 50.2,
"mem_last": "55.81MiB",
"samples": 9
},
"k6": {
"cpu_median": 27.2,
"cpu_max": 41.4,
"mem_last": "80.62MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 9.2,
"cpu_max": 27.2,
"mem_last": "12.93MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 46.8,
"cpu_max": 86.8,
"mem_last": "151.5MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 2.6,
"cpu_max": 3.0,
"mem_last": "3.855MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 7822,
"pool_wait_ms_avg": 33.443,
"pool_waiting_max": 37,
"pool_usage_ms_avg": 41.52,
"api_cpu_s": 25.41,
"intervals": 40
},
"pg": {
"xact_commit": 23276,
"xact_rollback": 0,
"tup_inserted": 11460,
"blks_read": 0,
"blks_hit": 253791,
"wal_records": 80896,
"wal_bytes": 7165147
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5826,
"total_ms": 1934.2,
"mean_ms": 0.332,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5826,
"total_ms": 871.9,
"mean_ms": 0.1497,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5826,
"total_ms": 646.7,
"mean_ms": 0.111,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11652,
"total_ms": 326.6,
"mean_ms": 0.028,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5826,
"total_ms": 187.0,
"mean_ms": 0.0321,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11805,
"total_ms": 19.6,
"mean_ms": 0.0017,
"query": "BEGIN"
},
{
"calls": 11805,
"total_ms": 18.7,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 153,
"total_ms": 0.8,
"mean_ms": 0.0055,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0612,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1098,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0278,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.01,
"statements_per_ok": 15.17,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4826,
"ok": 4826,
"dropped": 175,
"ok_rps": 241.3,
"error_ratio": 0,
"ok_p50_ms": 471.09206,
"ok_p95_ms": 1036.04970075,
"ok_p99_ms": 1227.1330874999999,
"vus_max": 355,
"cpu": {
"api_total": {
"cpu_median": 194.8,
"cpu_max": 202.4,
"samples": 9
},
"api-1": {
"cpu_median": 48.9,
"cpu_max": 51.7,
"mem_last": "69.15MiB",
"samples": 9
},
"api-2": {
"cpu_median": 50.1,
"cpu_max": 52.7,
"mem_last": "69.7MiB",
"samples": 9
},
"api-3": {
"cpu_median": 49.2,
"cpu_max": 50.9,
"mem_last": "58.52MiB",
"samples": 9
},
"api-4": {
"cpu_median": 46.6,
"cpu_max": 49.6,
"mem_last": "69.51MiB",
"samples": 9
},
"k6": {
"cpu_median": 31.9,
"cpu_max": 37.2,
"mem_last": "125.1MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 12.9,
"cpu_max": 24.0,
"mem_last": "14.38MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 58.1,
"cpu_max": 75.9,
"mem_last": "155.9MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 3.2,
"cpu_max": 3.7,
"mem_last": "3.793MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 9741,
"pool_wait_ms_avg": 136.726,
"pool_waiting_max": 60,
"pool_usage_ms_avg": 48.893,
"api_cpu_s": 29.49,
"intervals": 40
},
"pg": {
"xact_commit": 28905,
"xact_rollback": 0,
"tup_inserted": 14278,
"blks_read": 0,
"blks_hit": 327776,
"wal_records": 101018,
"wal_bytes": 8922212
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 4,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7178,
"total_ms": 2326.9,
"mean_ms": 0.3242,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7178,
"total_ms": 1012.0,
"mean_ms": 0.141,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 7178,
"total_ms": 940.5,
"mean_ms": 0.131,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 14356,
"total_ms": 372.6,
"mean_ms": 0.026,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7178,
"total_ms": 226.7,
"mean_ms": 0.0316,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14489,
"total_ms": 22.9,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 14489,
"total_ms": 19.5,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 133,
"total_ms": 0.9,
"mean_ms": 0.0068,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.13,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0413,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.023,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.99,
"statements_per_ok": 14.96,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 5001,
"ok": 4648,
"dropped": 1000,
"ok_rps": 232.4,
"error_ratio": 0.07059,
"ok_p50_ms": 2071.4172965,
"ok_p95_ms": 3384.998418,
"ok_p99_ms": 3670.734451479989,
"vus_max": 914,
"cpu": {
"api_total": {
"cpu_median": 191.1,
"cpu_max": 202.4,
"samples": 10
},
"api-1": {
"cpu_median": 49.3,
"cpu_max": 51.3,
"mem_last": "74.52MiB",
"samples": 10
},
"api-2": {
"cpu_median": 48.7,
"cpu_max": 51.6,
"mem_last": "62.03MiB",
"samples": 10
},
"api-3": {
"cpu_median": 48.2,
"cpu_max": 52.9,
"mem_last": "62.71MiB",
"samples": 10
},
"api-4": {
"cpu_median": 48.1,
"cpu_max": 52.3,
"mem_last": "61.5MiB",
"samples": 10
},
"k6": {
"cpu_median": 35.3,
"cpu_max": 63.6,
"mem_last": "286.9MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 14.5,
"cpu_max": 24.2,
"mem_last": "21.95MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 52.9,
"cpu_max": 83.8,
"mem_last": "161.9MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 7.7,
"cpu_max": 11.3,
"mem_last": "6.949MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 9134,
"pool_wait_ms_avg": 828.004,
"pool_waiting_max": 191,
"pool_usage_ms_avg": 58.025,
"api_cpu_s": 28.96,
"intervals": 40
},
"pg": {
"xact_commit": 29588,
"xact_rollback": 0,
"tup_inserted": 14514,
"blks_read": 0,
"blks_hit": 341907,
"wal_records": 104148,
"wal_bytes": 9315455
},
"lb": {
"chkdown_delta": 11,
"qmax": 310,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7338,
"total_ms": 10621.6,
"mean_ms": 1.4475,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 7338,
"total_ms": 2377.4,
"mean_ms": 0.324,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7338,
"total_ms": 1071.6,
"mean_ms": 0.146,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 14676,
"total_ms": 372.5,
"mean_ms": 0.0254,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7338,
"total_ms": 225.8,
"mean_ms": 0.0308,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14811,
"total_ms": 22.6,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 14811,
"total_ms": 19.1,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 135,
"total_ms": 0.8,
"mean_ms": 0.0057,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1021,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0488,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0322,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.37,
"statements_per_ok": 15.88,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 4529,
"ok": 4223,
"dropped": 2472,
"ok_rps": 211.2,
"error_ratio": 0.06756,
"ok_p50_ms": 2711.492793,
"ok_p95_ms": 4256.762477699999,
"ok_p99_ms": 4758.999461619997,
"vus_max": 1200,
"cpu": {
"api_total": {
"cpu_median": 187.2,
"cpu_max": 200.1,
"samples": 9
},
"api-1": {
"cpu_median": 48.0,
"cpu_max": 51.8,
"mem_last": "73.59MiB",
"samples": 9
},
"api-2": {
"cpu_median": 48.2,
"cpu_max": 51.2,
"mem_last": "75.34MiB",
"samples": 9
},
"api-3": {
"cpu_median": 48.9,
"cpu_max": 50.9,
"mem_last": "74.73MiB",
"samples": 9
},
"api-4": {
"cpu_median": 47.4,
"cpu_max": 51.4,
"mem_last": "63.69MiB",
"samples": 9
},
"k6": {
"cpu_median": 33.2,
"cpu_max": 77.0,
"mem_last": "360.6MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 11.4,
"cpu_max": 23.7,
"mem_last": "24.31MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 50.0,
"cpu_max": 73.4,
"mem_last": "166.2MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 7.5,
"cpu_max": 8.1,
"mem_last": "5.746MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 8639,
"pool_wait_ms_avg": 983.827,
"pool_waiting_max": 186,
"pool_usage_ms_avg": 60.319,
"api_cpu_s": 27.72,
"intervals": 40
},
"pg": {
"xact_commit": 27516,
"xact_rollback": 0,
"tup_inserted": 13466,
"blks_read": 0,
"blks_hit": 351837,
"wal_records": 98345,
"wal_bytes": 8479924
},
"lb": {
"chkdown_delta": 16,
"qmax": 372,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 5,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6912,
"total_ms": 12655.8,
"mean_ms": 1.831,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 6912,
"total_ms": 2437.3,
"mean_ms": 0.3526,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6912,
"total_ms": 1024.3,
"mean_ms": 0.1482,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 13824,
"total_ms": 354.5,
"mean_ms": 0.0256,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6912,
"total_ms": 218.7,
"mean_ms": 0.0316,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13961,
"total_ms": 21.7,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 13961,
"total_ms": 19.9,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 137,
"total_ms": 0.9,
"mean_ms": 0.0063,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1183,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0437,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0284,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.52,
"statements_per_ok": 16.47,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2001,
"ok": 2001,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 10.300163,
"ok_p95_ms": 81.956885,
"ok_p99_ms": 120.026972,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 119.8,
"cpu_max": 173.8,
"samples": 10
},
"api-1": {
"cpu_median": 31.3,
"cpu_max": 43.3,
"mem_last": "64.88MiB",
"samples": 10
},
"api-2": {
"cpu_median": 30.2,
"cpu_max": 44.3,
"mem_last": "66.55MiB",
"samples": 10
},
"api-3": {
"cpu_median": 24.1,
"cpu_max": 48.8,
"mem_last": "67.93MiB",
"samples": 10
},
"api-4": {
"cpu_median": 29.8,
"cpu_max": 44.4,
"mem_last": "65.21MiB",
"samples": 10
},
"k6": {
"cpu_median": 12.9,
"cpu_max": 13.8,
"mem_last": "26.08MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 4.6,
"cpu_max": 19.8,
"mem_last": "18.85MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 21.3,
"cpu_max": 26.0,
"mem_last": "93.49MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.4,
"cpu_max": 1.7,
"mem_last": "3.418MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 4117,
"pool_wait_ms_avg": 0.029,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 5.683,
"api_cpu_s": 11.23,
"intervals": 40
},
"pg": {
"xact_commit": 12312,
"xact_rollback": 0,
"tup_inserted": 6055,
"blks_read": 60,
"blks_hit": 134537,
"wal_records": 42946,
"wal_bytes": 4188796
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3002,
"total_ms": 729.6,
"mean_ms": 0.243,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3002,
"total_ms": 322.7,
"mean_ms": 0.1075,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 3002,
"total_ms": 193.5,
"mean_ms": 0.0645,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 6004,
"total_ms": 151.9,
"mean_ms": 0.0253,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3002,
"total_ms": 62.9,
"mean_ms": 0.021,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6181,
"total_ms": 9.0,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 6181,
"total_ms": 8.0,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 177,
"total_ms": 0.9,
"mean_ms": 0.0051,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.1677,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0492,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0296,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.15,
"statements_per_ok": 15.27,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2991,
"ok": 2991,
"dropped": 10,
"ok_rps": 149.6,
"error_ratio": 0,
"ok_p50_ms": 16.08078,
"ok_p95_ms": 229.2717,
"ok_p99_ms": 372.51721279999964,
"vus_max": 61,
"cpu": {
"api_total": {
"cpu_median": 142.7,
"cpu_max": 177.6,
"samples": 10
},
"api-1": {
"cpu_median": 37.0,
"cpu_max": 49.8,
"mem_last": "55.59MiB",
"samples": 10
},
"api-2": {
"cpu_median": 37.9,
"cpu_max": 51.1,
"mem_last": "67.61MiB",
"samples": 10
},
"api-3": {
"cpu_median": 35.8,
"cpu_max": 53.1,
"mem_last": "55.53MiB",
"samples": 10
},
"api-4": {
"cpu_median": 38.3,
"cpu_max": 51.9,
"mem_last": "66.28MiB",
"samples": 10
},
"k6": {
"cpu_median": 20.1,
"cpu_max": 22.0,
"mem_last": "32.14MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.4,
"cpu_max": 18.5,
"mem_last": "11.84MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.5,
"cpu_max": 47.6,
"mem_last": "147.9MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.1,
"cpu_max": 2.4,
"mem_last": "3.449MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6088,
"pool_wait_ms_avg": 0.226,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 12.748,
"api_cpu_s": 17.85,
"intervals": 40
},
"pg": {
"xact_commit": 18011,
"xact_rollback": 0,
"tup_inserted": 8824,
"blks_read": 0,
"blks_hit": 205504,
"wal_records": 62487,
"wal_bytes": 5541590
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4479,
"total_ms": 1303.9,
"mean_ms": 0.2911,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4479,
"total_ms": 590.3,
"mean_ms": 0.1318,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 4479,
"total_ms": 329.9,
"mean_ms": 0.0737,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8958,
"total_ms": 226.0,
"mean_ms": 0.0252,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4479,
"total_ms": 110.1,
"mean_ms": 0.0246,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9125,
"total_ms": 13.7,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 9125,
"total_ms": 13.4,
"mean_ms": 0.0015,
"query": "BEGIN"
},
{
"calls": 167,
"total_ms": 0.9,
"mean_ms": 0.0056,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.2,
"mean_ms": 0.154,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0441,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.043,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.02,
"statements_per_ok": 15.14,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3934,
"ok": 3934,
"dropped": 67,
"ok_rps": 196.7,
"error_ratio": 0,
"ok_p50_ms": 58.535457,
"ok_p95_ms": 506.05825574999994,
"ok_p99_ms": 735.6083260100002,
"vus_max": 176,
"cpu": {
"api_total": {
"cpu_median": 180.2,
"cpu_max": 191.3,
"samples": 9
},
"api-1": {
"cpu_median": 46.0,
"cpu_max": 49.2,
"mem_last": "55.46MiB",
"samples": 9
},
"api-2": {
"cpu_median": 44.4,
"cpu_max": 51.6,
"mem_last": "57.17MiB",
"samples": 9
},
"api-3": {
"cpu_median": 41.4,
"cpu_max": 51.2,
"mem_last": "55.71MiB",
"samples": 9
},
"api-4": {
"cpu_median": 41.1,
"cpu_max": 51.0,
"mem_last": "55.93MiB",
"samples": 9
},
"k6": {
"cpu_median": 24.8,
"cpu_max": 29.4,
"mem_last": "70.42MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 9.7,
"cpu_max": 22.1,
"mem_last": "12.83MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 52.4,
"cpu_max": 70.7,
"mem_last": "152MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 6.8,
"cpu_max": 9.5,
"mem_last": "3.266MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 7964,
"pool_wait_ms_avg": 16.316,
"pool_waiting_max": 19,
"pool_usage_ms_avg": 29.741,
"api_cpu_s": 23.93,
"intervals": 40
},
"pg": {
"xact_commit": 23723,
"xact_rollback": 0,
"tup_inserted": 11690,
"blks_read": 0,
"blks_hit": 252135,
"wal_records": 82485,
"wal_bytes": 7453571
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5862,
"total_ms": 1728.1,
"mean_ms": 0.2948,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5862,
"total_ms": 757.6,
"mean_ms": 0.1292,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5862,
"total_ms": 632.2,
"mean_ms": 0.1078,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11724,
"total_ms": 295.3,
"mean_ms": 0.0252,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5862,
"total_ms": 158.3,
"mean_ms": 0.027,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11878,
"total_ms": 18.5,
"mean_ms": 0.0016,
"query": "BEGIN"
},
{
"calls": 11878,
"total_ms": 17.1,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 154,
"total_ms": 0.8,
"mean_ms": 0.0054,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1108,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0503,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0218,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.03,
"statements_per_ok": 15.02,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4737,
"ok": 4737,
"dropped": 263,
"ok_rps": 236.8,
"error_ratio": 0,
"ok_p50_ms": 516.140248,
"ok_p95_ms": 1132.0239331999999,
"ok_p99_ms": 1435.51157744,
"vus_max": 428,
"cpu": {
"api_total": {
"cpu_median": 201.3,
"cpu_max": 205.7,
"samples": 9
},
"api-1": {
"cpu_median": 49.9,
"cpu_max": 51.8,
"mem_last": "56.47MiB",
"samples": 9
},
"api-2": {
"cpu_median": 50.0,
"cpu_max": 54.2,
"mem_last": "58.9MiB",
"samples": 9
},
"api-3": {
"cpu_median": 49.4,
"cpu_max": 51.6,
"mem_last": "56.29MiB",
"samples": 9
},
"api-4": {
"cpu_median": 51.0,
"cpu_max": 52.0,
"mem_last": "57.51MiB",
"samples": 9
},
"k6": {
"cpu_median": 34.3,
"cpu_max": 54.1,
"mem_last": "145.4MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 13.4,
"cpu_max": 22.3,
"mem_last": "15.32MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 62.3,
"cpu_max": 92.7,
"mem_last": "156.1MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 7.3,
"cpu_max": 8.7,
"mem_last": "3.898MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 9616,
"pool_wait_ms_avg": 157.497,
"pool_waiting_max": 57,
"pool_usage_ms_avg": 51.664,
"api_cpu_s": 29.41,
"intervals": 40
},
"pg": {
"xact_commit": 28477,
"xact_rollback": 0,
"tup_inserted": 14078,
"blks_read": 0,
"blks_hit": 314694,
"wal_records": 99577,
"wal_bytes": 8745233
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7102,
"total_ms": 2370.5,
"mean_ms": 0.3338,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7102,
"total_ms": 1041.5,
"mean_ms": 0.1467,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 7102,
"total_ms": 934.5,
"mean_ms": 0.1316,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 14204,
"total_ms": 355.1,
"mean_ms": 0.025,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7102,
"total_ms": 203.8,
"mean_ms": 0.0287,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14348,
"total_ms": 22.5,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 14348,
"total_ms": 20.8,
"mean_ms": 0.0015,
"query": "BEGIN"
},
{
"calls": 144,
"total_ms": 0.8,
"mean_ms": 0.0054,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1215,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0458,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0231,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.01,
"statements_per_ok": 15.08,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 4924,
"ok": 4497,
"dropped": 1077,
"ok_rps": 224.8,
"error_ratio": 0.08672,
"ok_p50_ms": 2188.955951,
"ok_p95_ms": 3537.9950315999995,
"ok_p99_ms": 3715.61750072,
"vus_max": 950,
"cpu": {
"api_total": {
"cpu_median": 191.7,
"cpu_max": 205.3,
"samples": 9
},
"api-1": {
"cpu_median": 48.9,
"cpu_max": 52.8,
"mem_last": "71.59MiB",
"samples": 9
},
"api-2": {
"cpu_median": 50.0,
"cpu_max": 51.3,
"mem_last": "62.5MiB",
"samples": 9
},
"api-3": {
"cpu_median": 49.8,
"cpu_max": 53.4,
"mem_last": "72.26MiB",
"samples": 9
},
"api-4": {
"cpu_median": 50.5,
"cpu_max": 51.1,
"mem_last": "61.52MiB",
"samples": 9
},
"k6": {
"cpu_median": 36.8,
"cpu_max": 59.8,
"mem_last": "336.1MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 14.1,
"cpu_max": 26.0,
"mem_last": "22.27MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 57.8,
"cpu_max": 79.3,
"mem_last": "161.9MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 7.2,
"cpu_max": 10.2,
"mem_last": "8.531MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 8946,
"pool_wait_ms_avg": 876.374,
"pool_waiting_max": 188,
"pool_usage_ms_avg": 58.138,
"api_cpu_s": 29.28,
"intervals": 40
},
"pg": {
"xact_commit": 28942,
"xact_rollback": 0,
"tup_inserted": 14188,
"blks_read": 0,
"blks_hit": 391113,
"wal_records": 102213,
"wal_bytes": 9054515
},
"lb": {
"chkdown_delta": 11,
"qmax": 292,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 1,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7143,
"total_ms": 9013.8,
"mean_ms": 1.2619,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 7143,
"total_ms": 2520.3,
"mean_ms": 0.3528,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7143,
"total_ms": 1065.6,
"mean_ms": 0.1492,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 14286,
"total_ms": 362.2,
"mean_ms": 0.0254,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7143,
"total_ms": 212.0,
"mean_ms": 0.0297,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14424,
"total_ms": 21.3,
"mean_ms": 0.0015,
"query": "BEGIN"
},
{
"calls": 14424,
"total_ms": 21.2,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 138,
"total_ms": 0.8,
"mean_ms": 0.0057,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0589,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1127,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.041,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.44,
"statements_per_ok": 15.98,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 4597,
"ok": 4103,
"dropped": 2404,
"ok_rps": 205.2,
"error_ratio": 0.10746,
"ok_p50_ms": 2748.361139,
"ok_p95_ms": 4117.3053162,
"ok_p99_ms": 4748.36108194,
"vus_max": 1200,
"cpu": {
"api_total": {
"cpu_median": 194.3,
"cpu_max": 202.5,
"samples": 10
},
"api-1": {
"cpu_median": 50.4,
"cpu_max": 52.0,
"mem_last": "75.15MiB",
"samples": 10
},
"api-2": {
"cpu_median": 49.0,
"cpu_max": 51.0,
"mem_last": "75.84MiB",
"samples": 10
},
"api-3": {
"cpu_median": 49.3,
"cpu_max": 51.3,
"mem_last": "75.96MiB",
"samples": 10
},
"api-4": {
"cpu_median": 49.0,
"cpu_max": 51.0,
"mem_last": "74.37MiB",
"samples": 10
},
"k6": {
"cpu_median": 36.1,
"cpu_max": 57.3,
"mem_last": "380.2MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 11.8,
"cpu_max": 25.0,
"mem_last": "23.97MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 47.9,
"cpu_max": 69.6,
"mem_last": "166.3MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.5,
"cpu_max": 5.1,
"mem_last": "6.324MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8150,
"pool_wait_ms_avg": 952.797,
"pool_waiting_max": 190,
"pool_usage_ms_avg": 58.304,
"api_cpu_s": 26.29,
"intervals": 39
},
"pg": {
"xact_commit": 27382,
"xact_rollback": 0,
"tup_inserted": 13546,
"blks_read": 0,
"blks_hit": 327386,
"wal_records": 97846,
"wal_bytes": 8479639
},
"lb": {
"chkdown_delta": 15,
"qmax": 298,
"up_after": 4,
"econ_delta": 344,
"eresp_delta": 4,
"redispatch_delta": 0
},
"statements": [
{
"calls": 6793,
"total_ms": 9740.1,
"mean_ms": 1.4338,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 6793,
"total_ms": 2188.9,
"mean_ms": 0.3222,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 6793,
"total_ms": 1006.4,
"mean_ms": 0.1482,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 13586,
"total_ms": 339.7,
"mean_ms": 0.025,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 6793,
"total_ms": 205.0,
"mean_ms": 0.0302,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 13715,
"total_ms": 21.8,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 13715,
"total_ms": 18.7,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 129,
"total_ms": 0.7,
"mean_ms": 0.0055,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1318,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0554,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.023,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.67,
"statements_per_ok": 16.65,
"repeat": 2,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 2000,
"ok": 2000,
"dropped": 0,
"ok_rps": 100.0,
"error_ratio": 0,
"ok_p50_ms": 11.3166325,
"ok_p95_ms": 79.14762619999999,
"ok_p99_ms": 117.88027991999995,
"vus_max": 40,
"cpu": {
"api_total": {
"cpu_median": 100.8,
"cpu_max": 185.0,
"samples": 10
},
"api-1": {
"cpu_median": 26.6,
"cpu_max": 51.6,
"mem_last": "65.68MiB",
"samples": 10
},
"api-2": {
"cpu_median": 31.9,
"cpu_max": 51.7,
"mem_last": "67.51MiB",
"samples": 10
},
"api-3": {
"cpu_median": 30.2,
"cpu_max": 51.5,
"mem_last": "67.78MiB",
"samples": 10
},
"api-4": {
"cpu_median": 25.2,
"cpu_max": 50.3,
"mem_last": "55.45MiB",
"samples": 10
},
"k6": {
"cpu_median": 12.9,
"cpu_max": 16.5,
"mem_last": "26.02MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 4.7,
"cpu_max": 19.8,
"mem_last": "11.94MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 29.4,
"cpu_max": 40.2,
"mem_last": "89.05MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.5,
"cpu_max": 2.7,
"mem_last": "3.691MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 4111,
"pool_wait_ms_avg": 0.006,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 6.304,
"api_cpu_s": 11.73,
"intervals": 40
},
"pg": {
"xact_commit": 12260,
"xact_rollback": 0,
"tup_inserted": 6056,
"blks_read": 60,
"blks_hit": 134341,
"wal_records": 42611,
"wal_bytes": 3989746
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 3001,
"total_ms": 760.6,
"mean_ms": 0.2534,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 3001,
"total_ms": 325.9,
"mean_ms": 0.1086,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 3001,
"total_ms": 192.4,
"mean_ms": 0.0641,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 6002,
"total_ms": 153.0,
"mean_ms": 0.0255,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 3001,
"total_ms": 61.2,
"mean_ms": 0.0204,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6177,
"total_ms": 8.5,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 6177,
"total_ms": 8.2,
"mean_ms": 0.0013,
"query": "COMMIT"
},
{
"calls": 175,
"total_ms": 0.9,
"mean_ms": 0.0054,
"query": "SELECT $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0738,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1443,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.028,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.13,
"statements_per_ok": 15.27,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2992,
"ok": 2992,
"dropped": 9,
"ok_rps": 149.6,
"error_ratio": 0,
"ok_p50_ms": 13.053227,
"ok_p95_ms": 143.46426579999994,
"ok_p99_ms": 212.89255586000033,
"vus_max": 59,
"cpu": {
"api_total": {
"cpu_median": 157.4,
"cpu_max": 200.0,
"samples": 10
},
"api-1": {
"cpu_median": 39.6,
"cpu_max": 49.8,
"mem_last": "55.68MiB",
"samples": 10
},
"api-2": {
"cpu_median": 42.6,
"cpu_max": 51.3,
"mem_last": "56.79MiB",
"samples": 10
},
"api-3": {
"cpu_median": 35.5,
"cpu_max": 48.1,
"mem_last": "56.86MiB",
"samples": 10
},
"api-4": {
"cpu_median": 30.8,
"cpu_max": 51.3,
"mem_last": "55MiB",
"samples": 10
},
"k6": {
"cpu_median": 19.7,
"cpu_max": 22.9,
"mem_last": "31.52MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.1,
"cpu_max": 17.6,
"mem_last": "11.91MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 40.8,
"cpu_max": 50.8,
"mem_last": "141.1MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 2.0,
"cpu_max": 2.5,
"mem_last": "3.707MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6110,
"pool_wait_ms_avg": 0.002,
"pool_waiting_max": 0,
"pool_usage_ms_avg": 10.893,
"api_cpu_s": 16.88,
"intervals": 40
},
"pg": {
"xact_commit": 18079,
"xact_rollback": 0,
"tup_inserted": 8856,
"blks_read": 0,
"blks_hit": 205209,
"wal_records": 62659,
"wal_bytes": 5577996
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4483,
"total_ms": 1240.0,
"mean_ms": 0.2766,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4483,
"total_ms": 553.4,
"mean_ms": 0.1235,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 4483,
"total_ms": 309.4,
"mean_ms": 0.069,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 8966,
"total_ms": 231.2,
"mean_ms": 0.0258,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4483,
"total_ms": 106.7,
"mean_ms": 0.0238,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 9136,
"total_ms": 12.9,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 9136,
"total_ms": 12.5,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 170,
"total_ms": 1.0,
"mean_ms": 0.006,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1197,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0586,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0231,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.04,
"statements_per_ok": 15.16,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3937,
"ok": 3937,
"dropped": 64,
"ok_rps": 196.8,
"error_ratio": 0,
"ok_p50_ms": 73.831768,
"ok_p95_ms": 446.8380135999997,
"ok_p99_ms": 554.06480936,
"vus_max": 139,
"cpu": {
"api_total": {
"cpu_median": 181.3,
"cpu_max": 198.5,
"samples": 9
},
"api-1": {
"cpu_median": 48.0,
"cpu_max": 51.9,
"mem_last": "68.5MiB",
"samples": 9
},
"api-2": {
"cpu_median": 47.4,
"cpu_max": 52.3,
"mem_last": "69.44MiB",
"samples": 9
},
"api-3": {
"cpu_median": 47.0,
"cpu_max": 51.3,
"mem_last": "69.87MiB",
"samples": 9
},
"api-4": {
"cpu_median": 41.5,
"cpu_max": 47.8,
"mem_last": "56.02MiB",
"samples": 9
},
"k6": {
"cpu_median": 27.2,
"cpu_max": 33.0,
"mem_last": "57.41MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 9.3,
"cpu_max": 23.1,
"mem_last": "12.52MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 47.0,
"cpu_max": 59.5,
"mem_last": "151.9MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 6.4,
"cpu_max": 8.3,
"mem_last": "3.707MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 7988,
"pool_wait_ms_avg": 10.874,
"pool_waiting_max": 14,
"pool_usage_ms_avg": 30.864,
"api_cpu_s": 23.66,
"intervals": 40
},
"pg": {
"xact_commit": 23669,
"xact_rollback": 0,
"tup_inserted": 11658,
"blks_read": 0,
"blks_hit": 253190,
"wal_records": 82258,
"wal_bytes": 7270127
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5900,
"total_ms": 1793.6,
"mean_ms": 0.304,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5900,
"total_ms": 768.3,
"mean_ms": 0.1302,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5900,
"total_ms": 567.0,
"mean_ms": 0.0961,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 11800,
"total_ms": 301.4,
"mean_ms": 0.0255,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5900,
"total_ms": 164.8,
"mean_ms": 0.0279,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 11961,
"total_ms": 17.3,
"mean_ms": 0.0014,
"query": "COMMIT"
},
{
"calls": 11961,
"total_ms": 16.3,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 161,
"total_ms": 0.8,
"mean_ms": 0.0053,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1111,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0554,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0234,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.01,
"statements_per_ok": 15.11,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 250,
"duration_s": 20,
"offered": 5000,
"requests": 4774,
"ok": 4774,
"dropped": 227,
"ok_rps": 238.7,
"error_ratio": 0,
"ok_p50_ms": 556.239102,
"ok_p95_ms": 1354.30700265,
"ok_p99_ms": 1603.5373057999968,
"vus_max": 387,
"cpu": {
"api_total": {
"cpu_median": 188.8,
"cpu_max": 196.0,
"samples": 9
},
"api-1": {
"cpu_median": 47.5,
"cpu_max": 50.2,
"mem_last": "57.85MiB",
"samples": 9
},
"api-2": {
"cpu_median": 47.0,
"cpu_max": 50.5,
"mem_last": "61.02MiB",
"samples": 9
},
"api-3": {
"cpu_median": 46.8,
"cpu_max": 49.1,
"mem_last": "70.6MiB",
"samples": 9
},
"api-4": {
"cpu_median": 46.3,
"cpu_max": 50.2,
"mem_last": "56.57MiB",
"samples": 9
},
"k6": {
"cpu_median": 32.9,
"cpu_max": 42.9,
"mem_last": "121.8MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 12.3,
"cpu_max": 25.9,
"mem_last": "15.23MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 60.1,
"cpu_max": 98.9,
"mem_last": "156.5MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 7.0,
"cpu_max": 8.7,
"mem_last": "4.051MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 9499,
"pool_wait_ms_avg": 188.954,
"pool_waiting_max": 81,
"pool_usage_ms_avg": 49.299,
"api_cpu_s": 28.93,
"intervals": 39
},
"pg": {
"xact_commit": 28718,
"xact_rollback": 0,
"tup_inserted": 14152,
"blks_read": 0,
"blks_hit": 316161,
"wal_records": 100505,
"wal_bytes": 8990420
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 5,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7134,
"total_ms": 2336.1,
"mean_ms": 0.3275,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7134,
"total_ms": 1025.8,
"mean_ms": 0.1438,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 7134,
"total_ms": 1019.5,
"mean_ms": 0.1429,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 14268,
"total_ms": 363.1,
"mean_ms": 0.0255,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7134,
"total_ms": 229.9,
"mean_ms": 0.0322,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14406,
"total_ms": 21.6,
"mean_ms": 0.0015,
"query": "COMMIT"
},
{
"calls": 14406,
"total_ms": 18.9,
"mean_ms": 0.0013,
"query": "BEGIN"
},
{
"calls": 138,
"total_ms": 0.7,
"mean_ms": 0.0052,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0962,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0391,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0239,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.02,
"statements_per_ok": 15.03,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 300,
"duration_s": 20,
"offered": 6000,
"requests": 4933,
"ok": 4544,
"dropped": 1068,
"ok_rps": 227.2,
"error_ratio": 0.07886,
"ok_p50_ms": 2346.2813260000003,
"ok_p95_ms": 3231.1088844999986,
"ok_p99_ms": 3388.2599049799996,
"vus_max": 949,
"cpu": {
"api_total": {
"cpu_median": 192.9,
"cpu_max": 203.1,
"samples": 10
},
"api-1": {
"cpu_median": 49.0,
"cpu_max": 50.3,
"mem_last": "61.19MiB",
"samples": 10
},
"api-2": {
"cpu_median": 48.1,
"cpu_max": 51.7,
"mem_last": "62.46MiB",
"samples": 10
},
"api-3": {
"cpu_median": 48.8,
"cpu_max": 50.3,
"mem_last": "74.53MiB",
"samples": 10
},
"api-4": {
"cpu_median": 48.4,
"cpu_max": 53.3,
"mem_last": "61.86MiB",
"samples": 10
},
"k6": {
"cpu_median": 34.6,
"cpu_max": 55.0,
"mem_last": "305.4MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 12.9,
"cpu_max": 23.1,
"mem_last": "21.45MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 52.1,
"cpu_max": 60.0,
"mem_last": "161.9MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 7.3,
"cpu_max": 9.4,
"mem_last": "5.961MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8773,
"pool_wait_ms_avg": 875.738,
"pool_waiting_max": 191,
"pool_usage_ms_avg": 57.914,
"api_cpu_s": 28.3,
"intervals": 40
},
"pg": {
"xact_commit": 29158,
"xact_rollback": 0,
"tup_inserted": 14358,
"blks_read": 0,
"blks_hit": 325327,
"wal_records": 103054,
"wal_bytes": 9101091
},
"lb": {
"chkdown_delta": 9,
"qmax": 200,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7202,
"total_ms": 10006.0,
"mean_ms": 1.3893,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 7202,
"total_ms": 2305.8,
"mean_ms": 0.3202,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7202,
"total_ms": 1061.6,
"mean_ms": 0.1474,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 14404,
"total_ms": 357.2,
"mean_ms": 0.0248,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7202,
"total_ms": 227.7,
"mean_ms": 0.0316,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14547,
"total_ms": 23.1,
"mean_ms": 0.0016,
"query": "COMMIT"
},
{
"calls": 14547,
"total_ms": 19.6,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 143,
"total_ms": 0.8,
"mean_ms": 0.0054,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1158,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0435,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0231,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.42,
"statements_per_ok": 15.94,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,210 @@
{
"config": "n4-api0.5-pg2-pool10",
"rate": 350,
"duration_s": 20,
"offered": 7000,
"requests": 4551,
"ok": 4309,
"dropped": 2450,
"ok_rps": 215.4,
"error_ratio": 0.05318,
"ok_p50_ms": 2596.842043,
"ok_p95_ms": 4211.608060799999,
"ok_p99_ms": 4912.60305484,
"vus_max": 1200,
"cpu": {
"api_total": {
"cpu_median": 184.5,
"cpu_max": 193.8,
"samples": 10
},
"api-1": {
"cpu_median": 47.8,
"cpu_max": 50.8,
"mem_last": "63.17MiB",
"samples": 10
},
"api-2": {
"cpu_median": 47.5,
"cpu_max": 50.4,
"mem_last": "75.74MiB",
"samples": 10
},
"api-3": {
"cpu_median": 48.8,
"cpu_max": 50.3,
"mem_last": "63.63MiB",
"samples": 10
},
"api-4": {
"cpu_median": 48.0,
"cpu_max": 50.5,
"mem_last": "76.49MiB",
"samples": 10
},
"k6": {
"cpu_median": 30.0,
"cpu_max": 47.5,
"mem_last": "377.9MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 13.5,
"cpu_max": 23.3,
"mem_last": "23.38MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 49.8,
"cpu_max": 74.1,
"mem_last": "166.7MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 7.1,
"cpu_max": 8.7,
"mem_last": "6.301MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 8587,
"pool_wait_ms_avg": 1022.183,
"pool_waiting_max": 190,
"pool_usage_ms_avg": 59.353,
"api_cpu_s": 27.39,
"intervals": 40
},
"pg": {
"xact_commit": 28511,
"xact_rollback": 0,
"tup_inserted": 13962,
"blks_read": 0,
"blks_hit": 358177,
"wal_records": 101881,
"wal_bytes": 8837592
},
"lb": {
"chkdown_delta": 16,
"qmax": 263,
"up_after": 4,
"econ_delta": 0,
"eresp_delta": 2,
"redispatch_delta": 0
},
"statements": [
{
"calls": 7118,
"total_ms": 12627.6,
"mean_ms": 1.774,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 7118,
"total_ms": 2408.9,
"mean_ms": 0.3384,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 7118,
"total_ms": 1045.2,
"mean_ms": 0.1468,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 14236,
"total_ms": 351.0,
"mean_ms": 0.0247,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 7118,
"total_ms": 212.0,
"mean_ms": 0.0298,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 14373,
"total_ms": 24.0,
"mean_ms": 0.0017,
"query": "COMMIT"
},
{
"calls": 14373,
"total_ms": 19.6,
"mean_ms": 0.0014,
"query": "BEGIN"
},
{
"calls": 137,
"total_ms": 0.7,
"mean_ms": 0.0054,
"query": "SELECT 1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0628,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.0953,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0231,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.62,
"statements_per_ok": 16.62,
"repeat": 3,
"limits": {
"api-1": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-2": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-3": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"api-4": {
"cpus": 0.5,
"memory_mib": 192,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 2.0,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,55 @@
config,repeat,rate,ok_rps,replicas_running,api_cpus_limit_total,api_mem_mib_each,pg_cpus_limit,requests,ok,dropped,error_ratio,p50_ms,p95_ms,p99_ms,api_cpu_pct_total,pg_cpu_pct,lb_cpu_pct,valkey_cpu_pct,k6_cpu_pct,pool_wait_ms_avg,pool_waiting_max,pg_commits_per_ok,statements_per_ok,lb_chkdown,lb_qmax,wal_bytes
n1-api2-pg2-pool10,1,100,100.0,1,2.0,768,2.0,2000,2000,0,0,7.5202495,28.052189000000002,98.54706085999999,75.4,25.6,4.3,1.6,12.2,0.008,0,6.05,15.08,0,0,4257360
n1-api2-pg2-pool10,1,150,149.6,1,2.0,768,2.0,2992,2992,9,0,8.871513,118.67201719999996,171.77328262000003,78.6,27.4,5.8,1.8,17.3,1.152,8,5.97,15.05,0,0,5541231
n1-api2-pg2-pool10,1,200,198.7,1,2.0,768,2.0,3973,3973,27,0,13.108659,141.66078979999992,229.97273719999976,103.1,38.4,7.6,3.7,21.8,5.268,22,5.98,15.03,0,0,7400028
n1-api2-pg2-pool10,1,250,209.0,1,2.0,768,2.0,4180,4180,820,0,2151.489912,2612.4429757,2703.39118036,110.6,40.1,9.6,5.9,28.9,469.808,189,6.06,15.35,0,300,7783781
n1-api2-pg2-pool10,1,300,204.3,1,2.0,768,2.0,4086,4086,1914,0,2793.8822490000002,3124.0045007500003,3211.33611675,99.0,46.8,9.6,2.1,27.4,461.578,190,6.55,16.1,0,400,8654533
n1-api2-pg2-pool10,1,350,196.0,1,2.0,768,2.0,3920,3920,3081,0,2988.8524905,3335.9800228999998,3396.08057728,100.7,39.7,8.9,3.0,27.4,469.138,187,6.57,16.38,0,418,7797623
n2-api1-pg2-pool10,1,100,100.0,2,2.0,384,2.0,2001,2001,0,0,7.80253,31.97001,59.668254,83.4,21.0,4.1,5.2,12.1,0.001,0,6.05,15.14,0,0,4214586
n2-api1-pg2-pool10,1,150,149.7,2,2.0,384,2.0,2993,2993,8,0,10.136685,84.46405700000001,245.7217186799998,121.3,39.0,6.6,5.6,19.1,0.528,1,5.98,15.08,0,0,5584015
n2-api1-pg2-pool10,1,200,197.6,2,2.0,384,2.0,3952,3952,49,0,15.75066,307.54845785,356.5846853,151.1,52.6,10.1,6.5,26.8,8.965,16,6.03,15.01,0,0,7386866
n2-api1-pg2-pool10,1,250,242.4,2,2.0,384,2.0,4848,4848,152,0,61.385312,630.4609458499998,819.2782447999998,173.5,53.4,11.7,7.4,32.4,38.644,53,5.92,15.0,0,0,9071845
n2-api1-pg2-pool10,1,300,281.4,2,2.0,384,2.0,5629,5629,372,0,785.742226,1390.0188632,1561.0473299599998,197.3,62.9,13.3,7.8,36.9,306.747,178,5.98,14.89,0,0,10236889
n2-api1-pg2-pool10,1,350,271.1,2,2.0,384,2.0,5423,5423,1577,0,1980.052642,2614.7839925999997,2785.22204186,196.2,65.1,15.6,3.0,44.3,692.084,190,6.13,15.64,0,200,10573711
n4-api0.5-pg2-pool10,1,100,100.0,4,2.0,192,2.0,2001,2001,0,0,11.298931,77.948327,125.443179,110.8,21.0,4.6,1.5,12.4,0.0,0,6.18,15.28,0,0,4239661
n4-api0.5-pg2-pool10,1,150,147.2,4,2.0,192,2.0,2945,2945,56,0,22.235219,460.2908489999996,789.53975552,148.2,40.3,7.4,6.0,19.9,5.766,4,6.04,15.22,0,0,5482545
n4-api0.5-pg2-pool10,1,200,193.6,4,2.0,192,2.0,3872,3872,129,0,180.1144295,751.6838958,982.1881199099998,190.4,46.8,9.2,2.6,27.2,33.443,37,6.01,15.17,0,0,7165147
n4-api0.5-pg2-pool10,1,250,241.3,4,2.0,192,2.0,4826,4826,175,0,471.09206,1036.04970075,1227.1330874999999,194.8,58.1,12.9,3.2,31.9,136.726,60,5.99,14.96,0,0,8922212
n4-api0.5-pg2-pool10,1,300,232.4,4,2.0,192,2.0,5001,4648,1000,0.07059,2071.4172965,3384.998418,3670.734451479989,191.1,52.9,14.5,7.7,35.3,828.004,191,6.37,15.88,11,310,9315455
n4-api0.5-pg2-pool10,1,350,211.2,4,2.0,192,2.0,4529,4223,2472,0.06756,2711.492793,4256.762477699999,4758.999461619997,187.2,50.0,11.4,7.5,33.2,983.827,186,6.52,16.47,16,372,8479924
n1-api2-pg2-pool10,2,100,100.0,1,2.0,768,2.0,2001,2001,0,0,7.250932,15.74861,37.804146,69.4,26.2,4.3,1.4,11.9,0.004,0,5.97,15.07,0,0,4237390
n1-api2-pg2-pool10,2,150,149.8,1,2.0,768,2.0,2995,2995,6,0,8.864475,65.56181429999967,149.6798073,87.1,26.0,5.7,1.8,17.5,0.601,0,5.99,15.04,0,0,5575852
n1-api2-pg2-pool10,2,200,198.5,1,2.0,768,2.0,3970,3970,30,0,22.8694485,200.6632921,247.88514344999996,101.6,45.4,7.8,6.0,23.0,9.836,27,5.95,15.04,0,0,7402176
n1-api2-pg2-pool10,2,250,213.9,1,2.0,768,2.0,4278,4278,723,0,2058.8974980000003,2484.172225,2522.18200123,99.4,49.1,9.4,2.2,28.7,456.896,186,6.07,15.23,0,300,7941124
n1-api2-pg2-pool10,2,300,200.3,1,2.0,768,2.0,4007,4007,1994,0,2824.94653,3517.4384534,3643.0758680599997,96.6,44.6,8.9,2.2,26.5,460.307,188,6.52,16.14,0,400,8578256
n1-api2-pg2-pool10,2,350,201.1,1,2.0,768,2.0,4021,4021,2979,0,2877.468829,3297.091424,3325.204687,97.7,40.9,9.3,2.2,27.6,460.264,186,6.47,16.16,0,416,7815846
n2-api1-pg2-pool10,2,100,100.0,2,2.0,384,2.0,2000,2000,0,0,8.537046,28.476515,43.90845024,86.2,19.9,4.3,1.5,11.7,0.0,0,6.05,15.15,0,0,4205568
n2-api1-pg2-pool10,2,150,150.1,2,2.0,384,2.0,3001,3001,0,0,9.827698,61.054989,102.130968,93.4,29.6,6.2,5.7,18.7,0.002,0,5.98,15.03,0,0,5537639
n2-api1-pg2-pool10,2,200,198.5,2,2.0,384,2.0,3970,3970,31,0,11.526496,198.70631769999994,318.32055527,138.3,42.9,9.0,4.9,26.9,5.568,22,6.0,14.98,0,0,7405507
n2-api1-pg2-pool10,2,250,246.1,2,2.0,384,2.0,4922,4922,79,0,20.5787295,382.6286063499999,488.18310679,163.3,54.7,11.2,7.4,32.1,25.254,46,5.99,15.02,0,0,9290226
n2-api1-pg2-pool10,2,300,290.6,2,2.0,384,2.0,5813,5813,188,0,344.621426,703.3058816,966.31137016,196.3,80.7,14.9,3.6,38.4,129.1,86,5.92,15.01,0,0,10557801
n2-api1-pg2-pool10,2,350,290.6,2,2.0,384,2.0,5813,5813,1187,0,1844.997341,2348.4341228000003,2418.85193552,197.9,65.6,14.2,3.5,42.0,610.592,188,6.15,15.23,0,200,11093891
n4-api0.5-pg2-pool10,2,100,100.0,4,2.0,192,2.0,2001,2001,0,0,10.300163,81.956885,120.026972,119.8,21.3,4.6,1.4,12.9,0.029,0,6.15,15.27,0,0,4188796
n4-api0.5-pg2-pool10,2,150,149.6,4,2.0,192,2.0,2991,2991,10,0,16.08078,229.2717,372.51721279999964,142.7,40.5,7.4,2.1,20.1,0.226,0,6.02,15.14,0,0,5541590
n4-api0.5-pg2-pool10,2,200,196.7,4,2.0,192,2.0,3934,3934,67,0,58.535457,506.05825574999994,735.6083260100002,180.2,52.4,9.7,6.8,24.8,16.316,19,6.03,15.02,0,0,7453571
n4-api0.5-pg2-pool10,2,250,236.8,4,2.0,192,2.0,4737,4737,263,0,516.140248,1132.0239331999999,1435.51157744,201.3,62.3,13.4,7.3,34.3,157.497,57,6.01,15.08,0,0,8745233
n4-api0.5-pg2-pool10,2,300,224.8,4,2.0,192,2.0,4924,4497,1077,0.08672,2188.955951,3537.9950315999995,3715.61750072,191.7,57.8,14.1,7.2,36.8,876.374,188,6.44,15.98,11,292,9054515
n4-api0.5-pg2-pool10,2,350,205.2,4,2.0,192,2.0,4597,4103,2404,0.10746,2748.361139,4117.3053162,4748.36108194,194.3,47.9,11.8,2.5,36.1,952.797,190,6.67,16.65,15,298,8479639
n1-api2-pg2-pool10,3,100,100.0,1,2.0,768,2.0,2001,2001,0,0,7.97685,15.575488,28.970835,62.3,26.8,4.3,1.4,11.7,0.0,0,6.01,15.08,0,0,4221770
n1-api2-pg2-pool10,3,150,149.8,1,2.0,768,2.0,2997,2997,4,0,8.971883,64.57286139999982,112.78668699999997,83.8,34.0,5.9,1.8,17.7,0.674,2,5.92,15.04,0,0,5534011
n1-api2-pg2-pool10,3,200,198.6,1,2.0,768,2.0,3971,3971,30,0,14.164644,155.16024099999998,257.88778990000014,103.2,45.1,7.9,6.1,22.7,5.877,26,5.95,14.97,0,0,7352333
n1-api2-pg2-pool10,3,250,219.8,1,2.0,768,2.0,4396,4396,604,0,1792.2371604999998,2446.1017865,2502.0942151000004,100.6,47.1,9.4,5.8,28.7,446.111,188,6.02,15.15,0,300,8301132
n1-api2-pg2-pool10,3,300,210.1,1,2.0,768,2.0,4201,4201,1800,0,2729.610138,2994.333622,3070.33987,105.5,46.6,9.4,5.9,27.2,444.989,187,6.43,15.91,0,400,8588339
n1-api2-pg2-pool10,3,350,198.8,1,2.0,768,2.0,3975,3975,3026,0,2965.12569,3324.3593067,3363.43487784,115.7,40.4,9.0,2.1,26.5,467.16,188,6.5,16.25,0,411,7895331
n2-api1-pg2-pool10,3,100,100.0,2,2.0,384,2.0,2001,2001,0,0,6.980185,25.244786,42.937627,77.5,18.7,4.2,4.6,11.4,0.0,0,6.08,15.15,0,0,4247596
n2-api1-pg2-pool10,3,150,149.2,2,2.0,384,2.0,2984,2984,17,0,8.575164000000001,96.4983846999999,263.7259766300002,109.1,36.9,6.2,2.3,16.6,0.335,2,6.01,15.09,0,0,5577453
n2-api1-pg2-pool10,3,200,198.5,2,2.0,384,2.0,3970,3970,30,0,11.589998000000001,162.31980174999998,253.8206895199997,110.8,40.9,8.8,2.6,24.9,1.351,6,6.0,15.07,0,0,7393664
n2-api1-pg2-pool10,3,250,245.1,2,2.0,384,2.0,4901,4901,100,0,38.657426,447.950927,561.220036,153.7,56.5,11.0,3.3,33.5,32.021,53,5.96,15.02,0,0,9182750
n2-api1-pg2-pool10,3,300,285.9,2,2.0,384,2.0,5718,5718,283,0,641.7616855,1123.39635615,1214.89919999,196.1,66.5,13.6,3.5,35.8,252.69,165,5.94,15.05,0,0,10415770
n2-api1-pg2-pool10,3,350,288.9,2,2.0,384,2.0,5779,5779,1221,0,1792.498706,2449.1132878999997,2635.6998854000003,201.3,70.2,13.8,8.0,42.6,637.12,189,6.16,15.25,0,200,11082104
n4-api0.5-pg2-pool10,3,100,100.0,4,2.0,192,2.0,2000,2000,0,0,11.3166325,79.14762619999999,117.88027991999995,100.8,29.4,4.7,1.5,12.9,0.006,0,6.13,15.27,0,0,3989746
n4-api0.5-pg2-pool10,3,150,149.6,4,2.0,192,2.0,2992,2992,9,0,13.053227,143.46426579999994,212.89255586000033,157.4,40.8,7.1,2.0,19.7,0.002,0,6.04,15.16,0,0,5577996
n4-api0.5-pg2-pool10,3,200,196.8,4,2.0,192,2.0,3937,3937,64,0,73.831768,446.8380135999997,554.06480936,181.3,47.0,9.3,6.4,27.2,10.874,14,6.01,15.11,0,0,7270127
n4-api0.5-pg2-pool10,3,250,238.7,4,2.0,192,2.0,4774,4774,227,0,556.239102,1354.30700265,1603.5373057999968,188.8,60.1,12.3,7.0,32.9,188.954,81,6.02,15.03,0,0,8990420
n4-api0.5-pg2-pool10,3,300,227.2,4,2.0,192,2.0,4933,4544,1068,0.07886,2346.2813260000003,3231.1088844999986,3388.2599049799996,192.9,52.1,12.9,7.3,34.6,875.738,191,6.42,15.94,9,200,9101091
n4-api0.5-pg2-pool10,3,350,215.4,4,2.0,192,2.0,4551,4309,2450,0.05318,2596.842043,4211.608060799999,4912.60305484,184.5,49.8,13.5,7.1,30.0,1022.183,190,6.62,16.62,16,263,8837592
1 config repeat rate ok_rps replicas_running api_cpus_limit_total api_mem_mib_each pg_cpus_limit requests ok dropped error_ratio p50_ms p95_ms p99_ms api_cpu_pct_total pg_cpu_pct lb_cpu_pct valkey_cpu_pct k6_cpu_pct pool_wait_ms_avg pool_waiting_max pg_commits_per_ok statements_per_ok lb_chkdown lb_qmax wal_bytes
2 n1-api2-pg2-pool10 1 100 100.0 1 2.0 768 2.0 2000 2000 0 0 7.5202495 28.052189000000002 98.54706085999999 75.4 25.6 4.3 1.6 12.2 0.008 0 6.05 15.08 0 0 4257360
3 n1-api2-pg2-pool10 1 150 149.6 1 2.0 768 2.0 2992 2992 9 0 8.871513 118.67201719999996 171.77328262000003 78.6 27.4 5.8 1.8 17.3 1.152 8 5.97 15.05 0 0 5541231
4 n1-api2-pg2-pool10 1 200 198.7 1 2.0 768 2.0 3973 3973 27 0 13.108659 141.66078979999992 229.97273719999976 103.1 38.4 7.6 3.7 21.8 5.268 22 5.98 15.03 0 0 7400028
5 n1-api2-pg2-pool10 1 250 209.0 1 2.0 768 2.0 4180 4180 820 0 2151.489912 2612.4429757 2703.39118036 110.6 40.1 9.6 5.9 28.9 469.808 189 6.06 15.35 0 300 7783781
6 n1-api2-pg2-pool10 1 300 204.3 1 2.0 768 2.0 4086 4086 1914 0 2793.8822490000002 3124.0045007500003 3211.33611675 99.0 46.8 9.6 2.1 27.4 461.578 190 6.55 16.1 0 400 8654533
7 n1-api2-pg2-pool10 1 350 196.0 1 2.0 768 2.0 3920 3920 3081 0 2988.8524905 3335.9800228999998 3396.08057728 100.7 39.7 8.9 3.0 27.4 469.138 187 6.57 16.38 0 418 7797623
8 n2-api1-pg2-pool10 1 100 100.0 2 2.0 384 2.0 2001 2001 0 0 7.80253 31.97001 59.668254 83.4 21.0 4.1 5.2 12.1 0.001 0 6.05 15.14 0 0 4214586
9 n2-api1-pg2-pool10 1 150 149.7 2 2.0 384 2.0 2993 2993 8 0 10.136685 84.46405700000001 245.7217186799998 121.3 39.0 6.6 5.6 19.1 0.528 1 5.98 15.08 0 0 5584015
10 n2-api1-pg2-pool10 1 200 197.6 2 2.0 384 2.0 3952 3952 49 0 15.75066 307.54845785 356.5846853 151.1 52.6 10.1 6.5 26.8 8.965 16 6.03 15.01 0 0 7386866
11 n2-api1-pg2-pool10 1 250 242.4 2 2.0 384 2.0 4848 4848 152 0 61.385312 630.4609458499998 819.2782447999998 173.5 53.4 11.7 7.4 32.4 38.644 53 5.92 15.0 0 0 9071845
12 n2-api1-pg2-pool10 1 300 281.4 2 2.0 384 2.0 5629 5629 372 0 785.742226 1390.0188632 1561.0473299599998 197.3 62.9 13.3 7.8 36.9 306.747 178 5.98 14.89 0 0 10236889
13 n2-api1-pg2-pool10 1 350 271.1 2 2.0 384 2.0 5423 5423 1577 0 1980.052642 2614.7839925999997 2785.22204186 196.2 65.1 15.6 3.0 44.3 692.084 190 6.13 15.64 0 200 10573711
14 n4-api0.5-pg2-pool10 1 100 100.0 4 2.0 192 2.0 2001 2001 0 0 11.298931 77.948327 125.443179 110.8 21.0 4.6 1.5 12.4 0.0 0 6.18 15.28 0 0 4239661
15 n4-api0.5-pg2-pool10 1 150 147.2 4 2.0 192 2.0 2945 2945 56 0 22.235219 460.2908489999996 789.53975552 148.2 40.3 7.4 6.0 19.9 5.766 4 6.04 15.22 0 0 5482545
16 n4-api0.5-pg2-pool10 1 200 193.6 4 2.0 192 2.0 3872 3872 129 0 180.1144295 751.6838958 982.1881199099998 190.4 46.8 9.2 2.6 27.2 33.443 37 6.01 15.17 0 0 7165147
17 n4-api0.5-pg2-pool10 1 250 241.3 4 2.0 192 2.0 4826 4826 175 0 471.09206 1036.04970075 1227.1330874999999 194.8 58.1 12.9 3.2 31.9 136.726 60 5.99 14.96 0 0 8922212
18 n4-api0.5-pg2-pool10 1 300 232.4 4 2.0 192 2.0 5001 4648 1000 0.07059 2071.4172965 3384.998418 3670.734451479989 191.1 52.9 14.5 7.7 35.3 828.004 191 6.37 15.88 11 310 9315455
19 n4-api0.5-pg2-pool10 1 350 211.2 4 2.0 192 2.0 4529 4223 2472 0.06756 2711.492793 4256.762477699999 4758.999461619997 187.2 50.0 11.4 7.5 33.2 983.827 186 6.52 16.47 16 372 8479924
20 n1-api2-pg2-pool10 2 100 100.0 1 2.0 768 2.0 2001 2001 0 0 7.250932 15.74861 37.804146 69.4 26.2 4.3 1.4 11.9 0.004 0 5.97 15.07 0 0 4237390
21 n1-api2-pg2-pool10 2 150 149.8 1 2.0 768 2.0 2995 2995 6 0 8.864475 65.56181429999967 149.6798073 87.1 26.0 5.7 1.8 17.5 0.601 0 5.99 15.04 0 0 5575852
22 n1-api2-pg2-pool10 2 200 198.5 1 2.0 768 2.0 3970 3970 30 0 22.8694485 200.6632921 247.88514344999996 101.6 45.4 7.8 6.0 23.0 9.836 27 5.95 15.04 0 0 7402176
23 n1-api2-pg2-pool10 2 250 213.9 1 2.0 768 2.0 4278 4278 723 0 2058.8974980000003 2484.172225 2522.18200123 99.4 49.1 9.4 2.2 28.7 456.896 186 6.07 15.23 0 300 7941124
24 n1-api2-pg2-pool10 2 300 200.3 1 2.0 768 2.0 4007 4007 1994 0 2824.94653 3517.4384534 3643.0758680599997 96.6 44.6 8.9 2.2 26.5 460.307 188 6.52 16.14 0 400 8578256
25 n1-api2-pg2-pool10 2 350 201.1 1 2.0 768 2.0 4021 4021 2979 0 2877.468829 3297.091424 3325.204687 97.7 40.9 9.3 2.2 27.6 460.264 186 6.47 16.16 0 416 7815846
26 n2-api1-pg2-pool10 2 100 100.0 2 2.0 384 2.0 2000 2000 0 0 8.537046 28.476515 43.90845024 86.2 19.9 4.3 1.5 11.7 0.0 0 6.05 15.15 0 0 4205568
27 n2-api1-pg2-pool10 2 150 150.1 2 2.0 384 2.0 3001 3001 0 0 9.827698 61.054989 102.130968 93.4 29.6 6.2 5.7 18.7 0.002 0 5.98 15.03 0 0 5537639
28 n2-api1-pg2-pool10 2 200 198.5 2 2.0 384 2.0 3970 3970 31 0 11.526496 198.70631769999994 318.32055527 138.3 42.9 9.0 4.9 26.9 5.568 22 6.0 14.98 0 0 7405507
29 n2-api1-pg2-pool10 2 250 246.1 2 2.0 384 2.0 4922 4922 79 0 20.5787295 382.6286063499999 488.18310679 163.3 54.7 11.2 7.4 32.1 25.254 46 5.99 15.02 0 0 9290226
30 n2-api1-pg2-pool10 2 300 290.6 2 2.0 384 2.0 5813 5813 188 0 344.621426 703.3058816 966.31137016 196.3 80.7 14.9 3.6 38.4 129.1 86 5.92 15.01 0 0 10557801
31 n2-api1-pg2-pool10 2 350 290.6 2 2.0 384 2.0 5813 5813 1187 0 1844.997341 2348.4341228000003 2418.85193552 197.9 65.6 14.2 3.5 42.0 610.592 188 6.15 15.23 0 200 11093891
32 n4-api0.5-pg2-pool10 2 100 100.0 4 2.0 192 2.0 2001 2001 0 0 10.300163 81.956885 120.026972 119.8 21.3 4.6 1.4 12.9 0.029 0 6.15 15.27 0 0 4188796
33 n4-api0.5-pg2-pool10 2 150 149.6 4 2.0 192 2.0 2991 2991 10 0 16.08078 229.2717 372.51721279999964 142.7 40.5 7.4 2.1 20.1 0.226 0 6.02 15.14 0 0 5541590
34 n4-api0.5-pg2-pool10 2 200 196.7 4 2.0 192 2.0 3934 3934 67 0 58.535457 506.05825574999994 735.6083260100002 180.2 52.4 9.7 6.8 24.8 16.316 19 6.03 15.02 0 0 7453571
35 n4-api0.5-pg2-pool10 2 250 236.8 4 2.0 192 2.0 4737 4737 263 0 516.140248 1132.0239331999999 1435.51157744 201.3 62.3 13.4 7.3 34.3 157.497 57 6.01 15.08 0 0 8745233
36 n4-api0.5-pg2-pool10 2 300 224.8 4 2.0 192 2.0 4924 4497 1077 0.08672 2188.955951 3537.9950315999995 3715.61750072 191.7 57.8 14.1 7.2 36.8 876.374 188 6.44 15.98 11 292 9054515
37 n4-api0.5-pg2-pool10 2 350 205.2 4 2.0 192 2.0 4597 4103 2404 0.10746 2748.361139 4117.3053162 4748.36108194 194.3 47.9 11.8 2.5 36.1 952.797 190 6.67 16.65 15 298 8479639
38 n1-api2-pg2-pool10 3 100 100.0 1 2.0 768 2.0 2001 2001 0 0 7.97685 15.575488 28.970835 62.3 26.8 4.3 1.4 11.7 0.0 0 6.01 15.08 0 0 4221770
39 n1-api2-pg2-pool10 3 150 149.8 1 2.0 768 2.0 2997 2997 4 0 8.971883 64.57286139999982 112.78668699999997 83.8 34.0 5.9 1.8 17.7 0.674 2 5.92 15.04 0 0 5534011
40 n1-api2-pg2-pool10 3 200 198.6 1 2.0 768 2.0 3971 3971 30 0 14.164644 155.16024099999998 257.88778990000014 103.2 45.1 7.9 6.1 22.7 5.877 26 5.95 14.97 0 0 7352333
41 n1-api2-pg2-pool10 3 250 219.8 1 2.0 768 2.0 4396 4396 604 0 1792.2371604999998 2446.1017865 2502.0942151000004 100.6 47.1 9.4 5.8 28.7 446.111 188 6.02 15.15 0 300 8301132
42 n1-api2-pg2-pool10 3 300 210.1 1 2.0 768 2.0 4201 4201 1800 0 2729.610138 2994.333622 3070.33987 105.5 46.6 9.4 5.9 27.2 444.989 187 6.43 15.91 0 400 8588339
43 n1-api2-pg2-pool10 3 350 198.8 1 2.0 768 2.0 3975 3975 3026 0 2965.12569 3324.3593067 3363.43487784 115.7 40.4 9.0 2.1 26.5 467.16 188 6.5 16.25 0 411 7895331
44 n2-api1-pg2-pool10 3 100 100.0 2 2.0 384 2.0 2001 2001 0 0 6.980185 25.244786 42.937627 77.5 18.7 4.2 4.6 11.4 0.0 0 6.08 15.15 0 0 4247596
45 n2-api1-pg2-pool10 3 150 149.2 2 2.0 384 2.0 2984 2984 17 0 8.575164000000001 96.4983846999999 263.7259766300002 109.1 36.9 6.2 2.3 16.6 0.335 2 6.01 15.09 0 0 5577453
46 n2-api1-pg2-pool10 3 200 198.5 2 2.0 384 2.0 3970 3970 30 0 11.589998000000001 162.31980174999998 253.8206895199997 110.8 40.9 8.8 2.6 24.9 1.351 6 6.0 15.07 0 0 7393664
47 n2-api1-pg2-pool10 3 250 245.1 2 2.0 384 2.0 4901 4901 100 0 38.657426 447.950927 561.220036 153.7 56.5 11.0 3.3 33.5 32.021 53 5.96 15.02 0 0 9182750
48 n2-api1-pg2-pool10 3 300 285.9 2 2.0 384 2.0 5718 5718 283 0 641.7616855 1123.39635615 1214.89919999 196.1 66.5 13.6 3.5 35.8 252.69 165 5.94 15.05 0 0 10415770
49 n2-api1-pg2-pool10 3 350 288.9 2 2.0 384 2.0 5779 5779 1221 0 1792.498706 2449.1132878999997 2635.6998854000003 201.3 70.2 13.8 8.0 42.6 637.12 189 6.16 15.25 0 200 11082104
50 n4-api0.5-pg2-pool10 3 100 100.0 4 2.0 192 2.0 2000 2000 0 0 11.3166325 79.14762619999999 117.88027991999995 100.8 29.4 4.7 1.5 12.9 0.006 0 6.13 15.27 0 0 3989746
51 n4-api0.5-pg2-pool10 3 150 149.6 4 2.0 192 2.0 2992 2992 9 0 13.053227 143.46426579999994 212.89255586000033 157.4 40.8 7.1 2.0 19.7 0.002 0 6.04 15.16 0 0 5577996
52 n4-api0.5-pg2-pool10 3 200 196.8 4 2.0 192 2.0 3937 3937 64 0 73.831768 446.8380135999997 554.06480936 181.3 47.0 9.3 6.4 27.2 10.874 14 6.01 15.11 0 0 7270127
53 n4-api0.5-pg2-pool10 3 250 238.7 4 2.0 192 2.0 4774 4774 227 0 556.239102 1354.30700265 1603.5373057999968 188.8 60.1 12.3 7.0 32.9 188.954 81 6.02 15.03 0 0 8990420
54 n4-api0.5-pg2-pool10 3 300 227.2 4 2.0 192 2.0 4933 4544 1068 0.07886 2346.2813260000003 3231.1088844999986 3388.2599049799996 192.9 52.1 12.9 7.3 34.6 875.738 191 6.42 15.94 9 200 9101091
55 n4-api0.5-pg2-pool10 3 350 215.4 4 2.0 192 2.0 4551 4309 2450 0.05318 2596.842043 4211.608060799999 4912.60305484 184.5 49.8 13.5 7.1 30.0 1022.183 190 6.62 16.62 16 263 8837592
@@ -0,0 +1,82 @@
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 760 420" font-family="system-ui, sans-serif" font-size="12">
<style>.bg{fill:#fcfcfb}.ink{fill:#0b0b0b}.ink2{fill:#52514e}.grid{stroke:#e4e3df}.ref{stroke:#b5b3ac}.l0{stroke:#2a78d6;fill:none}.d0{fill:#2a78d6;stroke:#fcfcfb}.l1{stroke:#eb6834;fill:none}.d1{fill:#eb6834;stroke:#fcfcfb}.l2{stroke:#1baf7a;fill:none}.d2{fill:#1baf7a;stroke:#fcfcfb}.l3{stroke:#eda100;fill:none}.d3{fill:#eda100;stroke:#fcfcfb}@media (prefers-color-scheme: dark){.bg{fill:#1a1a19}.ink{fill:#fff}.ink2{fill:#c3c2b7}.grid{stroke:#34332f}.ref{stroke:#6b6a64}.l0{stroke:#3987e5}.d0{fill:#3987e5;stroke:#1a1a19}.l1{stroke:#d95926}.d1{fill:#d95926;stroke:#1a1a19}.l2{stroke:#199e70}.d2{fill:#199e70;stroke:#1a1a19}.l3{stroke:#c98500}.d3{fill:#c98500;stroke:#1a1a19}}</style>
<rect class="bg" width="760" height="420"/>
<text class="ink" x="64" y="22" font-size="14" font-weight="600">A-fixed-total/baseline</text>
<line class="grid" x1="64" x2="550" y1="368.0" y2="368.0" stroke-width="1"/>
<text class="ink2" x="56" y="372.0" text-anchor="end">0</text>
<line class="grid" x1="64" x2="550" y1="325.4" y2="325.4" stroke-width="1"/>
<text class="ink2" x="56" y="329.4" text-anchor="end">50</text>
<line class="grid" x1="64" x2="550" y1="282.8" y2="282.8" stroke-width="1"/>
<text class="ink2" x="56" y="286.8" text-anchor="end">100</text>
<line class="grid" x1="64" x2="550" y1="240.2" y2="240.2" stroke-width="1"/>
<text class="ink2" x="56" y="244.2" text-anchor="end">150</text>
<line class="grid" x1="64" x2="550" y1="197.6" y2="197.6" stroke-width="1"/>
<text class="ink2" x="56" y="201.6" text-anchor="end">200</text>
<line class="grid" x1="64" x2="550" y1="155.0" y2="155.0" stroke-width="1"/>
<text class="ink2" x="56" y="159.0" text-anchor="end">250</text>
<line class="grid" x1="64" x2="550" y1="112.4" y2="112.4" stroke-width="1"/>
<text class="ink2" x="56" y="116.4" text-anchor="end">300</text>
<line class="grid" x1="64" x2="550" y1="69.8" y2="69.8" stroke-width="1"/>
<text class="ink2" x="56" y="73.8" text-anchor="end">350</text>
<text class="ink2" x="196.2" y="386" text-anchor="middle">100</text>
<text class="ink2" x="262.4" y="386" text-anchor="middle">150</text>
<text class="ink2" x="328.5" y="386" text-anchor="middle">200</text>
<text class="ink2" x="394.6" y="386" text-anchor="middle">250</text>
<text class="ink2" x="460.7" y="386" text-anchor="middle">300</text>
<text class="ink2" x="526.9" y="386" text-anchor="middle">350</text>
<text class="ink2" x="307.0" y="408" text-anchor="middle">предложенная нагрузка, запросов/с (open-loop)</text>
<text class="ink2" transform="translate(16 204.0) rotate(-90)" text-anchor="middle">успешных запросов/с (201)</text>
<line class="ref" x1="64.0" y1="368.0" x2="526.9" y2="69.8" stroke-width="1"/>
<text class="ink2" x="272.9" y="221.6" text-anchor="end">обслужено = предложено</text>
<polyline class="l0" points="196.2,282.8 262.4,240.4 328.5,198.8 394.6,185.8 460.7,193.9 526.9,198.6" stroke-width="2" stroke-linejoin="round" stroke-linecap="round"/>
<line class="l0" x1="196.2" x2="196.2" y1="282.8" y2="282.8" stroke-width="2" opacity="0.5"/>
<circle class="d0" cx="196.2" cy="282.8" r="4" stroke-width="2"><title>1 × 2 CPU API, БД 2 CPU: предложено 100/с, успешно 100/с (повторы 100–100), p95 16 мс</title></circle>
<line class="l0" x1="262.4" x2="262.4" y1="240.5" y2="240.4" stroke-width="2" opacity="0.5"/>
<circle class="d0" cx="262.4" cy="240.4" r="4" stroke-width="2"><title>1 × 2 CPU API, БД 2 CPU: предложено 150/с, успешно 149.8/с (повторы 149.6–149.8), p95 66 мс</title></circle>
<line class="l0" x1="328.5" x2="328.5" y1="198.9" y2="198.7" stroke-width="2" opacity="0.5"/>
<circle class="d0" cx="328.5" cy="198.8" r="4" stroke-width="2"><title>1 × 2 CPU API, БД 2 CPU: предложено 200/с, успешно 198.6/с (повторы 198.5–198.7), p95 155 мс</title></circle>
<line class="l0" x1="394.6" x2="394.6" y1="189.9" y2="180.7" stroke-width="2" opacity="0.5"/>
<circle class="d0" cx="394.6" cy="185.8" r="4" stroke-width="2"><title>1 × 2 CPU API, БД 2 CPU: предложено 250/с, успешно 213.9/с (повторы 209–219.8), p95 2484 мс</title></circle>
<line class="l0" x1="460.7" x2="460.7" y1="197.4" y2="189.0" stroke-width="2" opacity="0.5"/>
<circle class="d0" cx="460.7" cy="193.9" r="4" stroke-width="2"><title>1 × 2 CPU API, БД 2 CPU: предложено 300/с, успешно 204.3/с (повторы 200.3–210.1), p95 3124 мс</title></circle>
<line class="l0" x1="526.9" x2="526.9" y1="201.0" y2="196.7" stroke-width="2" opacity="0.5"/>
<circle class="d0" cx="526.9" cy="198.6" r="4" stroke-width="2"><title>1 × 2 CPU API, БД 2 CPU: предложено 350/с, успешно 198.8/с (повторы 196–201.1), p95 3324 мс</title></circle>
<text class="ink2" x="394.6" y="175.8" text-anchor="middle">max 214</text>
<line class="l0" x1="584" x2="602" y1="50" y2="50" stroke-width="2"/>
<text class="ink" x="608" y="54">1 × 2 CPU API, БД 2</text>
<text class="ink" x="608" y="68">CPU</text>
<polyline class="l1" points="196.2,282.8 262.4,240.5 328.5,198.9 394.6,159.2 460.7,124.4 526.9,121.9" stroke-width="2" stroke-linejoin="round" stroke-linecap="round"/>
<line class="l1" x1="196.2" x2="196.2" y1="282.8" y2="282.8" stroke-width="2" opacity="0.5"/>
<circle class="d1" cx="196.2" cy="282.8" r="4" stroke-width="2"><title>2 × 1 CPU API, БД 2 CPU: предложено 100/с, успешно 100/с (повторы 100–100), p95 28 мс</title></circle>
<line class="l1" x1="262.4" x2="262.4" y1="240.9" y2="240.1" stroke-width="2" opacity="0.5"/>
<circle class="d1" cx="262.4" cy="240.5" r="4" stroke-width="2"><title>2 × 1 CPU API, БД 2 CPU: предложено 150/с, успешно 149.7/с (повторы 149.2–150.1), p95 84 мс</title></circle>
<line class="l1" x1="328.5" x2="328.5" y1="199.7" y2="198.9" stroke-width="2" opacity="0.5"/>
<circle class="d1" cx="328.5" cy="198.9" r="4" stroke-width="2"><title>2 × 1 CPU API, БД 2 CPU: предложено 200/с, успешно 198.5/с (повторы 197.6–198.5), p95 199 мс</title></circle>
<line class="l1" x1="394.6" x2="394.6" y1="161.5" y2="158.3" stroke-width="2" opacity="0.5"/>
<circle class="d1" cx="394.6" cy="159.2" r="4" stroke-width="2"><title>2 × 1 CPU API, БД 2 CPU: предложено 250/с, успешно 245.1/с (повторы 242.4–246.1), p95 448 мс</title></circle>
<line class="l1" x1="460.7" x2="460.7" y1="128.3" y2="120.4" stroke-width="2" opacity="0.5"/>
<circle class="d1" cx="460.7" cy="124.4" r="4" stroke-width="2"><title>2 × 1 CPU API, БД 2 CPU: предложено 300/с, успешно 285.9/с (повторы 281.4–290.6), p95 1123 мс</title></circle>
<line class="l1" x1="526.9" x2="526.9" y1="137.0" y2="120.4" stroke-width="2" opacity="0.5"/>
<circle class="d1" cx="526.9" cy="121.9" r="4" stroke-width="2"><title>2 × 1 CPU API, БД 2 CPU: предложено 350/с, успешно 288.9/с (повторы 271.1–290.6), p95 2449 мс</title></circle>
<text class="ink2" x="526.9" y="111.9" text-anchor="middle">max 289</text>
<line class="l1" x1="584" x2="602" y1="90" y2="90" stroke-width="2"/>
<text class="ink" x="608" y="94">2 × 1 CPU API, БД 2</text>
<text class="ink" x="608" y="108">CPU</text>
<polyline class="l2" points="196.2,282.8 262.4,240.5 328.5,200.4 394.6,164.6 460.7,174.4 526.9,188.1" stroke-width="2" stroke-linejoin="round" stroke-linecap="round"/>
<line class="l2" x1="196.2" x2="196.2" y1="282.8" y2="282.8" stroke-width="2" opacity="0.5"/>
<circle class="d2" cx="196.2" cy="282.8" r="4" stroke-width="2"><title>4 × 0.5 CPU API, БД 2 CPU: предложено 100/с, успешно 100/с (повторы 100–100), p95 79 мс</title></circle>
<line class="l2" x1="262.4" x2="262.4" y1="242.6" y2="240.5" stroke-width="2" opacity="0.5"/>
<circle class="d2" cx="262.4" cy="240.5" r="4" stroke-width="2"><title>4 × 0.5 CPU API, БД 2 CPU: предложено 150/с, успешно 149.6/с (повторы 147.2–149.6), p95 229 мс</title></circle>
<line class="l2" x1="328.5" x2="328.5" y1="203.1" y2="200.3" stroke-width="2" opacity="0.5"/>
<circle class="d2" cx="328.5" cy="200.4" r="4" stroke-width="2"><title>4 × 0.5 CPU API, БД 2 CPU: предложено 200/с, успешно 196.7/с (повторы 193.6–196.8), p95 506 мс</title></circle>
<line class="l2" x1="394.6" x2="394.6" y1="166.3" y2="162.4" stroke-width="2" opacity="0.5"/>
<circle class="d2" cx="394.6" cy="164.6" r="4" stroke-width="2"><title>4 × 0.5 CPU API, БД 2 CPU: предложено 250/с, успешно 238.7/с (повторы 236.8–241.3), p95 1132 мс</title></circle>
<line class="l2" x1="460.7" x2="460.7" y1="176.5" y2="170.0" stroke-width="2" opacity="0.5"/>
<circle class="d2" cx="460.7" cy="174.4" r="4" stroke-width="2"><title>4 × 0.5 CPU API, БД 2 CPU: предложено 300/с, успешно 227.2/с (повторы 224.8–232.4), p95 3385 мс</title></circle>
<line class="l2" x1="526.9" x2="526.9" y1="193.2" y2="184.5" stroke-width="2" opacity="0.5"/>
<circle class="d2" cx="526.9" cy="188.1" r="4" stroke-width="2"><title>4 × 0.5 CPU API, БД 2 CPU: предложено 350/с, успешно 211.2/с (повторы 205.2–215.4), p95 4212 мс</title></circle>
<text class="ink2" x="394.6" y="154.6" text-anchor="middle">max 239</text>
<line class="l2" x1="584" x2="602" y1="130" y2="130" stroke-width="2"/>
<text class="ink" x="608" y="134">4 × 0.5 CPU API, БД 2</text>
<text class="ink" x="608" y="148">CPU</text>
</svg>

After

Width:  |  Height:  |  Size: 9.5 KiB

@@ -0,0 +1,66 @@
{
"started": "2026-10-19T11:39:47+00:00",
"measured_commit": "3f45eb1",
"image": "mtusi-messenger:bench-3f45eb1",
"bench_tooling_dirty": true,
"host": {
"os": "Windows 10",
"docker": "29.8.0 8 CPU 15686365184 B 6.18.54.1-microsoft-standard-WSL2"
},
"cpusets": "k6 0-1, lb+valkey 2, postgres 3-4, api 5-7 (bench/track1/compose.bench.yaml)",
"workload": "k6 constant-arrival-rate POST message, new Idempotency-Key, 64 chats, warmup 10s + measure 20s per step, fresh stand per config x repeat",
"rates": [
100,
150,
200,
250,
300,
350,
400
],
"repeats": 3,
"configs": [
{
"replicas": 1,
"api_cpus": 1.0,
"api_memory": "384m",
"pg_cpus": 0.5,
"pool": 10,
"env": {}
},
{
"replicas": 2,
"api_cpus": 1.0,
"api_memory": "384m",
"pg_cpus": 0.5,
"pool": 10,
"env": {}
},
{
"replicas": 3,
"api_cpus": 1.0,
"api_memory": "384m",
"pg_cpus": 0.5,
"pool": 10,
"env": {}
}
],
"args": [
"--git-ref",
"3f45eb1",
"--repeats",
"3",
"--series",
"B-replicas-vs-db",
"--label",
"baseline",
"--rates",
"100,150,200,250,300,350,400",
"--config",
"replicas=1,pg_cpus=0.5",
"--config",
"replicas=2,pg_cpus=0.5",
"--config",
"replicas=3,pg_cpus=0.5"
]
}
@@ -0,0 +1,174 @@
{
"config": "n1-api1-pg0.5-pool10",
"rate": 100,
"duration_s": 20,
"offered": 2000,
"requests": 1996,
"ok": 1996,
"dropped": 5,
"ok_rps": 99.8,
"error_ratio": 0,
"ok_p50_ms": 7.9635355,
"ok_p95_ms": 112.08113375,
"ok_p99_ms": 236.7766470499998,
"vus_max": 50,
"cpu": {
"api_total": {
"cpu_median": 66.9,
"cpu_max": 84.4,
"samples": 9
},
"api-1": {
"cpu_median": 66.9,
"cpu_max": 84.4,
"mem_last": "58.47MiB",
"samples": 9
},
"k6": {
"cpu_median": 11.5,
"cpu_max": 12.8,
"mem_last": "28.98MiB",
"samples": 9
},
"lb-1": {
"cpu_median": 4.3,
"cpu_max": 13.6,
"mem_last": "11.91MiB",
"samples": 9
},
"postgres-1": {
"cpu_median": 20.0,
"cpu_max": 31.5,
"mem_last": "70.87MiB",
"samples": 9
},
"valkey-1": {
"cpu_median": 4.8,
"cpu_max": 5.2,
"mem_last": "3.598MiB",
"samples": 9
}
},
"pool": {
"pool_requests": 4027,
"pool_wait_ms_avg": 0.835,
"pool_waiting_max": 8,
"pool_usage_ms_avg": 6.067,
"api_cpu_s": 9.41,
"intervals": 10
},
"pg": {
"xact_commit": 12072,
"xact_rollback": 0,
"tup_inserted": 6090,
"blks_read": 82,
"blks_hit": 147608,
"wal_records": 42920,
"wal_bytes": 4217677
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 2992,
"total_ms": 687.2,
"mean_ms": 0.2297,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 2992,
"total_ms": 292.9,
"mean_ms": 0.0979,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5984,
"total_ms": 218.8,
"mean_ms": 0.0366,
"query": "SELECT $3 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 2992,
"total_ms": 171.6,
"mean_ms": 0.0574,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval $2) WHERE id = $1 RE"
},
{
"calls": 2992,
"total_ms": 55.5,
"mean_ms": 0.0186,
"query": "SELECT $2 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 6029,
"total_ms": 6.7,
"mean_ms": 0.0011,
"query": "BEGIN"
},
{
"calls": 6029,
"total_ms": 6.6,
"mean_ms": 0.0011,
"query": "COMMIT"
},
{
"calls": 45,
"total_ms": 0.2,
"mean_ms": 0.0049,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1422,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0537,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0292,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.05,
"statements_per_ok": 15.06,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 0.5,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api1-pg0.5-pool10",
"rate": 150,
"duration_s": 20,
"offered": 3000,
"requests": 2924,
"ok": 2924,
"dropped": 76,
"ok_rps": 146.2,
"error_ratio": 0,
"ok_p50_ms": 101.757823,
"ok_p95_ms": 522.14559485,
"ok_p99_ms": 820.8025748599999,
"vus_max": 158,
"cpu": {
"api_total": {
"cpu_median": 88.7,
"cpu_max": 92.5,
"samples": 10
},
"api-1": {
"cpu_median": 88.7,
"cpu_max": 92.5,
"mem_last": "69.05MiB",
"samples": 10
},
"k6": {
"cpu_median": 19.8,
"cpu_max": 23.7,
"mem_last": "65.76MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 6.1,
"cpu_max": 17.2,
"mem_last": "12.8MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 30.3,
"cpu_max": 45.7,
"mem_last": "75.14MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 5.2,
"cpu_max": 8.0,
"mem_last": "3.715MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 5296,
"pool_wait_ms_avg": 46.406,
"pool_waiting_max": 32,
"pool_usage_ms_avg": 14.966,
"api_cpu_s": 12.54,
"intervals": 9
},
"pg": {
"xact_commit": 17470,
"xact_rollback": 0,
"tup_inserted": 8684,
"blks_read": 0,
"blks_hit": 203182,
"wal_records": 61567,
"wal_bytes": 5431690
},
"lb": {
"chkdown_delta": 0,
"qmax": 0,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 4379,
"total_ms": 982.0,
"mean_ms": 0.2242,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 4379,
"total_ms": 451.5,
"mean_ms": 0.1031,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 8758,
"total_ms": 291.5,
"mean_ms": 0.0333,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 4379,
"total_ms": 279.3,
"mean_ms": 0.0638,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 4379,
"total_ms": 102.9,
"mean_ms": 0.0235,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 8800,
"total_ms": 24.6,
"mean_ms": 0.0028,
"query": "BEGIN"
},
{
"calls": 8800,
"total_ms": 10.3,
"mean_ms": 0.0012,
"query": "COMMIT"
},
{
"calls": 42,
"total_ms": 0.2,
"mean_ms": 0.0054,
"query": "SELECT $1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1335,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0646,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0203,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 5.97,
"statements_per_ok": 15.02,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 0.5,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}
@@ -0,0 +1,174 @@
{
"config": "n1-api1-pg0.5-pool10",
"rate": 200,
"duration_s": 20,
"offered": 4000,
"requests": 3559,
"ok": 3559,
"dropped": 441,
"ok_rps": 177.9,
"error_ratio": 0,
"ok_p50_ms": 1446.033837,
"ok_p95_ms": 2403.2934942,
"ok_p99_ms": 2545.62725412,
"vus_max": 594,
"cpu": {
"api_total": {
"cpu_median": 89.9,
"cpu_max": 98.7,
"samples": 10
},
"api-1": {
"cpu_median": 89.9,
"cpu_max": 98.7,
"mem_last": "62.94MiB",
"samples": 10
},
"k6": {
"cpu_median": 22.9,
"cpu_max": 33.7,
"mem_last": "179.6MiB",
"samples": 10
},
"lb-1": {
"cpu_median": 7.8,
"cpu_max": 18.5,
"mem_last": "16.89MiB",
"samples": 10
},
"postgres-1": {
"cpu_median": 38.1,
"cpu_max": 43.9,
"mem_last": "82.46MiB",
"samples": 10
},
"valkey-1": {
"cpu_median": 1.9,
"cpu_max": 2.2,
"mem_last": "3.828MiB",
"samples": 10
}
},
"pool": {
"pool_requests": 6747,
"pool_wait_ms_avg": 484.946,
"pool_waiting_max": 191,
"pool_usage_ms_avg": 21.698,
"api_cpu_s": 16.46,
"intervals": 10
},
"pg": {
"xact_commit": 21414,
"xact_rollback": 0,
"tup_inserted": 10590,
"blks_read": 0,
"blks_hit": 255656,
"wal_records": 74991,
"wal_bytes": 6640417
},
"lb": {
"chkdown_delta": 0,
"qmax": 200,
"up_after": 1,
"econ_delta": 0,
"eresp_delta": 0,
"redispatch_delta": 0
},
"statements": [
{
"calls": 5380,
"total_ms": 1306.2,
"mean_ms": 0.2428,
"query": "INSERT INTO idempotency_keys (user_id, chat_id, key, request_hash, message_id) VALUES ($1, $2, $3, $4, $5) ON "
},
{
"calls": 5380,
"total_ms": 659.5,
"mean_ms": 0.1226,
"query": "INSERT INTO messages (id, chat_id, sender_id, text, created_at) VALUES ($1, $2, $3, $4, $5)"
},
{
"calls": 5380,
"total_ms": 375.7,
"mean_ms": 0.0698,
"query": "UPDATE chats SET last_message_at = greatest(clock_timestamp(), last_message_at + interval '1 microsecond') WHE"
},
{
"calls": 10760,
"total_ms": 335.5,
"mean_ms": 0.0312,
"query": "SELECT 1 FROM chat_members WHERE chat_id = $1 AND user_id = $2"
},
{
"calls": 5380,
"total_ms": 134.3,
"mean_ms": 0.025,
"query": "SELECT 1 FROM ONLY \"public\".\"messages\" x WHERE \"id\" OPERATOR(pg_catalog.=) $1 FOR KEY SHARE OF x"
},
{
"calls": 10797,
"total_ms": 11.5,
"mean_ms": 0.0011,
"query": "COMMIT"
},
{
"calls": 10797,
"total_ms": 10.1,
"mean_ms": 0.0009,
"query": "BEGIN"
},
{
"calls": 37,
"total_ms": 0.2,
"mean_ms": 0.0047,
"query": "SELECT 1"
},
{
"calls": 1,
"total_ms": 0.1,
"mean_ms": 0.1196,
"query": "SELECT pg_stat_statements_reset()"
},
{
"calls": 2,
"total_ms": 0.1,
"mean_ms": 0.0438,
"query": "SELECT xact_commit, xact_rollback, tup_inserted, blks_read, blks_hit FROM pg_stat_database WHERE datname = $1"
},
{
"calls": 2,
"total_ms": 0.0,
"mean_ms": 0.0223,
"query": "SELECT wal_records, wal_bytes FROM pg_stat_wal"
}
],
"pg_commits_per_ok": 6.02,
"statements_per_ok": 15.15,
"repeat": 1,
"limits": {
"api-1": {
"cpus": 1.0,
"memory_mib": 384,
"cpuset": "5-7",
"image": "mtusi-messenger:bench-3f45eb1"
},
"lb-1": {
"cpus": 1.0,
"memory_mib": 128,
"cpuset": "2",
"image": "haproxy:3.2.26-alpine@sha256:254e6370f01cd654e0744edf607c67f33777c1b6a9dc0d4b28bcba7171460122"
},
"postgres-1": {
"cpus": 0.5,
"memory_mib": 1024,
"cpuset": "3-4",
"image": "postgres:18.6-bookworm@sha256:afc7e2d441324c0388fa80c3d24f733b4194a4eb7f47dd8ee2b08eb1a24a647c"
},
"valkey-1": {
"cpus": 0.5,
"memory_mib": 256,
"cpuset": "2",
"image": "valkey/valkey:8.1.3-bookworm@sha256:fea8b3e67b15729d4bb70589eb03367bab9ad1ee89c876f54327fc7c6e618571"
}
}
}

Some files were not shown because too many files have changed in this diff Show More