diff --git a/.env.example b/.env.example index 5a6b7a9..b85ec28 100644 --- a/.env.example +++ b/.env.example @@ -12,6 +12,12 @@ CACHE_URL_FILE=/run/secrets/cache_url API_REPLICAS=2 # Per-replica CPU limit. Series A keeps replicas * API_CPU_LIMIT = 1.0. API_CPU_LIMIT=0.50 +# Per-replica memory. Track 1 keeps replicas * API_MEM_LIMIT = 512m. +API_MEM_LIMIT=256m +# Base profile for docker compose up. scripts/run_track1.py passes the same defaults. +HTTP_WORKERS=8 +DB_POOL_SIZE=8 +NGINX_WORKERS=1 # PostgreSQL cgroup CPU limit. DB_CPU_LIMIT=1.0 # Nginx cgroup CPU limit. 0.25 saturated in the track-1 plateau runs. diff --git a/REPORT.md b/REPORT.md index c742fc1..edba236 100644 --- a/REPORT.md +++ b/REPORT.md @@ -80,42 +80,46 @@ Nginx 1.28: `resolver 127.0.0.11 valid=2s`, `server api:8080 resolve`, `max_fail ## Трек 1. Цена масштабирования -Гипотеза. При суммарном CPU API = 1.0 дополнительные реплики не поднимают RPS: каждая реплика упирается в свой кусок квоты. Пока квота PostgreSQL заполнена, лишний CPU базы поднимает RPS; когда квота API и Nginx уже полные, дальнейший CPU базы даёт меньше. Пул соединений и число процессов Nginx сами по себе не добавляют пропускную способность, пока у реплики 8 потоков обработки и те же CPU-квоты. +Гипотеза. При суммарном CPU API = 1.0 и суммарной RAM API = 512 MiB дополнительные реплики не поднимают RPS: каждая реплика упирается в свой кусок квоты. Пока квота PostgreSQL заполнена, лишний CPU базы поднимает RPS; когда квота API и Nginx уже полные, дальнейший CPU базы даёт меньше. -Методика. Публичный `tests/load.py`, closed-loop `POST /messages`, `--duration 30 --concurrency 50`. Перед каждым зачётным окном отброшенный прогрев 10 с с теми же 50 воркерами. Каждый запуск создаёт свой чат. Ошибок записи во всех зачётных окнах ниже не было. `docker stats` снимался примерно раз в 3 с только во время 30 с; в выводе Docker 100% — это одно ядро. Лимиты после `up` проверены по `HostConfig.NanoCpus`. +Методика. Публичный `tests/load.py`, closed-loop `POST /messages`, `--duration 30 --concurrency 50`. Перед каждым зачётным окном отброшенный прогрев 10 с с теми же 50 воркерами. Каждый запуск создаёт свой чат. `docker stats` снимался примерно раз в 3 с только во время 30 с; в выводе Docker 100% — это одно ядро. Скрипт `scripts/run_track1.py` на каждую точку делает `docker compose down` и `docker compose up -d --build`, передаёт переменные окружения и до нагрузки сверяет `docker inspect`. -Фиксировано во всех точках серий A и B: `shared_buffers=256MB`, `synchronous_commit=off`, `max_connections=200`, память db 512 MiB, cache 0.5 CPU, Nginx `worker_processes 1` и квота lb 0.25 CPU, пул PostgreSQL 8, 8 потоков HTTP на реплику. Хост: Intel Core i5-12450H, 12 потоков, 14 GiB RAM. Скрипт прогона: `scripts/run_track1.py`. Сырые JSON лежат в `evidence/track1-*.json`. +Методика воспроизводима: суммарные ресурсы фиксировались через переменные окружения и проверялись через docker inspect. Базовые значения (8 потоков, пул 8, один процесс Nginx) восстановлены в коде по умолчанию. `HTTP_WORKERS` и `DB_POOL_SIZE` читаются процессом API, `NGINX_WORKERS` подставляется в `lb/nginx.conf.template`. Без этих переменных `docker compose up` поднимает тот же профиль 8/8/1 и 256 MiB на реплику. В сериях A и B суммарная RAM API всегда 512 MiB: 1×512m, 2×256m, 4×128m. Скрипт записывает прогон только если `HostConfig.NanoCpus` и `HostConfig.Memory` совпали с этими значениями, сумма Memory API равна 536870912, в логе каждой реплики есть `pg pool ready (8 connections)` и `http workers: 8`, а `nginx -T` показывает `worker_processes 1;`. -### Серия A. Реплики API, суммарный CPU = 1.0, CPU базы = 2.0 +Фиксировано во всех точках серий A и B: `shared_buffers=256MB`, `synchronous_commit=off`, `max_connections=200`, память db 512 MiB, cache 0.5 CPU, квота lb 0.25 CPU, пул PostgreSQL 8, 8 потоков HTTP на реплику, Nginx `worker_processes 1` и `worker_connections 1024`. Хост: Intel Core i5-12450H, 12 потоков, 14 GiB RAM. Сырые JSON: `evidence/track1-A-*.json` и `evidence/track1-B-*.json`. -| Реплики | CPU на реплику | Успешный RPS | p50, мс | p95, мс | p99, мс | Ошибки | -| --- | --- | --- | --- | --- | --- | --- | -| 1 | 1.0 | 349.994 | 121.454 | 202.572 | 1115.09 | 0 | -| 2 | 0.5 | 327.611 | 131.433 | 289.233 | 404.102 | 0 | -| 4 | 0.25 | 320.002 | 151.294 | 295.941 | 407.03 | 0 | +### Серия A. Реплики API, суммарный CPU = 1.0, суммарная RAM = 512 MiB, CPU базы = 2.0 -От 1 к 2 репликам RPS −6.4%, от 1 к 4 — −8.6%. Медиана стала хуже: 121 → 131 → 151 мс. На одной реплике API держал около 90–101% (квота 1 ядро), база — до 207% (квота 2 ядра), Nginx — до 26% (квота 0.25 ядра). На четырёх репликах максимум API по контейнеру был 27.5% при квоте 0.25 ядра. +| Реплики | CPU на реплику | RAM на реплику | Успешный RPS | p50, мс | p95, мс | p99, мс | Ошибки | +| --- | --- | --- | --- | --- | --- | --- | --- | +| 1 | 1.0 | 512 MiB | 378.703 | 90.288 | 204.67 | 1124.75 | 21 | +| 2 | 0.5 | 256 MiB | 346.098 | 128.325 | 269.126 | 380.157 | 0 | +| 4 | 0.25 | 128 MiB | 327.486 | 102.151 | 487.734 | 626.914 | 0 | + +Сумма RAM по inspect на каждой точке — 536870912 байт. NanoCpus реплик: 1000000000, 500000000 и 250000000. От 1 к 2 репликам RPS −8.6%, от 1 к 4 — −13.5%. На одной реплике API держал максимум 105% при квоте 1 ядро, база — 212% при квоте 2 ядра, Nginx — 28% при квоте 0.25 ядра. В этом окне 21 неуспешный запрос из 11649 (error ratio 0.0018); успешный RPS считается только по 11628 ответам. На четырёх репликах максимум API по контейнеру был 27.2% при квоте 0.25 ядра, ошибок не было. ![Серия A: RPS и p99 от числа реплик](evidence/api_scaling.png) -Рост числа реплик при том же суммарном CPU API не помогает. Запрос по-прежнему ограничен долей ядра на процесс, а дробление добавляет хвост задержки. p99 одной реплики (1115 мс) — один длинный хвост в том же окне, не отдельная серия. +Рост числа реплик при том же суммарном CPU и той же суммарной RAM API не помогает. Запрос по-прежнему ограничен долей ядра на процесс. p99 одной реплики (1125 мс) — длинный хвост в том же окне. -### Серия B. CPU PostgreSQL, API зафиксирован: 4 × 0.25 +### Серия B. CPU PostgreSQL, API зафиксирован: 4 × 0.25 CPU и 4 × 128 MiB | CPU базы | Успешный RPS | p50, мс | p95, мс | p99, мс | Ошибки | Кто упёрся в квоту | | --- | --- | --- | --- | --- | --- | --- | -| 1.0 | 165.641 | 297.372 | 499.711 | 598.3 | 0 | база, max 107% | -| 2.0 | 331.742 | 112.739 | 377.522 | 459.163 | 0 | база max 207%; API max 27.6%; lb max 26.0% | -| 4.0 | 417.528 | 99.961 | 288.263 | 371.702 | 0 | API max 27.4%; lb max 26.2%; база max 242% из 400% | +| 1.0 | 178.949 | 286.653 | 480.072 | 538.062 | 0 | база, max 103% | +| 2.0 | 334.227 | 117.124 | 366.356 | 455.007 | 0 | база max 191%; API max 28.2%; lb max 25.8% | +| 4.0 | 348.996 | 110.449 | 394.75 | 479.704 | 0 | API max 27.9%; lb max 26.0%; база max 204% из 400% | -Второе ядро базы относительно первого: 331.742 / 165.641 = **+100.3% RPS**. Это подтверждённое улучшение: при 1.0 CPU база была на квоте, API (max 14%) и Nginx (max 12%) — нет. Четвёртое ядро относительно двух: 417.528 / 331.742 − 1 = **+25.9%**. База при лимите 4.0 использовала максимум около 2.4 ядра, а API и Nginx уже стояли на своих квотах. Поэтому дальнейший CPU базы помогает меньше. +На всех трёх точках inspect показал 4 контейнера API по NanoCpus 250000000 и Memory 134217728, сумма 536870912. NanoCpus базы: 1000000000, 2000000000, 4000000000. + +Второе ядро базы относительно первого: 334.227 / 178.949 − 1 = **+86.8% RPS**. Это подтверждённое улучшение: при 1.0 CPU база была на квоте, API (max 19%) и Nginx (max 12%) — нет. Четвёртое ядро относительно двух: 348.996 / 334.227 − 1 = **+4.4%**. База при лимите 4.0 использовала максимум около 2.0 ядра, а API и Nginx уже стояли на своих квотах. Поэтому дальнейший CPU базы почти не поднимает RPS. ![Серия B: RPS и p99 от CPU базы](evidence/db_scaling.png) -Крестики и звёзды на графике — не точки этой серии. Это поздний парный прогон, он ниже исходных 417.528. См. следующий блок. - ### Тюнинг Nginx и пула +Строки ниже сняты до параметризации: память была 256 MiB на реплику, а пул и потоки были зашиты в бинарник. Текущий `scripts/run_track1.py` эти точки не повторяет. Зачётные серии — таблицы A и B выше. + Проверялось ровно то, что предлагалось как снятие узкого места: `worker_processes auto`, `worker_connections 4096`, пул `PgPool(..., 32)`. В первом таком прогоне потоков HTTP было 8, поэтому из 32 соединений реплика одновременно занимала не больше восьми. Сразу после серии B тот же тюнинг вместе с квотой lb 2.0 CPU дал 336.611 RPS. Повтор исходной точки серии B (пул 8, один процесс Nginx, lb 0.25, база 4.0 CPU), сделанный позже, дал уже не 417.528, а **344.846** RPS (p50 129.553, p95 307.781, p99 397.828). Абсолютные 417.528 в конце сессии не воспроизвелись: долгий прогон на одном ноутбуке смешал сравнение. Зачётный вывод по тюнингу — только парная пара, снятая подряд: @@ -130,7 +134,7 @@ Nginx 1.28: `resolver 127.0.0.11 valid=2s`, `server api:8080 resolve`, `max_fail Снятие потолка потоков (`workers_n` 8 → 32) при том же пуле 32, Nginx `auto`, 4 × 0.25 CPU API и 4.0 CPU базы: (273.854 − 350.038) / 350.038 = **−21.8% RPS**. p50 120.943 → 176.149 мс, p99 396.982 → 519.185 мс, ошибок 0. База при этом занята меньше (среднее 110% и максимум 179% против 130% и 204% при 8 потоках), то есть лишние соединения пула не начали работать. Квота API по-прежнему выбрана: максимум 27.6% при лимите 0.25 ядра на реплику. 32 потока делят четверть ядра и добавляют переключения и ожидание на мьютексах, а CPU, которым можно было бы держать больше запросов к базе, не появляется. Ограничение в 8 потоков не было причиной простоя пула. -Плато серии B снимается вторым ядром PostgreSQL (+100.3% от 165.641 до 331.742), а не пулом, не числом процессов Nginx и не числом потоков HTTP. Квота API в 1.0 CPU на все реплики при этом не менялась. +Плато воспроизводимой серии B снимается вторым ядром PostgreSQL (+86.8%, от 178.949 до 334.227). Квота API в 1.0 CPU на все реплики при этом не менялась. Более ранний тюнинг пула и числа процессов Nginx на незафиксированной RAM дал +1.5% и −21.8% и зачётным результатом не является. ## Ограничения и следующие шаги @@ -138,7 +142,7 @@ Nginx 1.28: `resolver 127.0.0.11 valid=2s`, `server api:8080 resolve`, `max_fail Nginx OSS не опрашивает `/health/ready` сам. Готовность для LB — это 502/503/504 и обрыв соединения плюс `max_fails`. Окно drain (2 с) реплика ещё может ответить 201 со своим `X-Instance-Id`. -`tests/load.py` — closed-loop, без прозрачных retry, процентили только успешных запросов. Серия трека 1 — отдельные окна по 30 с при concurrency 50; один ранний прогон на 36 с в таблице отказов к ней не относится. Поздний повтор точки «база 4.0 CPU» дал 344.846 RPS вместо 417.528, поэтому абсолютный уровень после долгой сессии плавает. +`tests/load.py` — closed-loop, без прозрачных retry, процентили только успешных запросов. Серия трека 1 — отдельные окна по 30 с при concurrency 50; один ранний прогон на 36 с в таблице отказов к ней не относится. В воспроизводимой серии A точка с одной репликой дала 21 неуспешный запрос; остальные пять окон серии A и B — без ошибок. Сессии живут в Valkey с `appendfsync everysec`: теоретически последняя секунда записи может не попасть на диск при жёстком kill cache. `restart.sh` шлёт обычный SIGTERM, persistence-тест с этим прошёл. diff --git a/compose.yaml b/compose.yaml index 9422f59..fa81339 100644 --- a/compose.yaml +++ b/compose.yaml @@ -50,6 +50,8 @@ services: CACHE_URL_FILE: /run/secrets/cache_url IDEMPOTENCY_TTL_SECONDS: ${IDEMPOTENCY_TTL_SECONDS:-86400} DRAIN_SECONDS: ${DRAIN_SECONDS:-2} + HTTP_WORKERS: ${HTTP_WORKERS:-8} + DB_POOL_SIZE: ${DB_POOL_SIZE:-8} secrets: - database_url - cache_url @@ -64,7 +66,7 @@ services: resources: limits: cpus: ${API_CPU_LIMIT:-0.50} - memory: ${API_MEM_LIMIT:-256M} + memory: ${API_MEM_LIMIT:-256m} cpus: ${API_CPU_LIMIT:-0.50} # для совместимости с docker compose up mem_limit: ${API_MEM_LIMIT:-256m} stop_grace_period: 15s @@ -76,8 +78,13 @@ services: condition: service_healthy ports: - "8080:8080" + environment: + NGINX_WORKERS: ${NGINX_WORKERS:-1} + NGINX_ENVSUBST_TEMPLATE_DIR: /etc/nginx/templates + NGINX_ENVSUBST_OUTPUT_DIR: /etc/nginx + NGINX_ENVSUBST_FILTER: "^NGINX_" volumes: - - ./lb/nginx.conf:/etc/nginx/nginx.conf:ro + - ./lb/nginx.conf.template:/etc/nginx/templates/nginx.conf.template:ro healthcheck: test: ["CMD", "wget", "-q", "-O", "/dev/null", "http://127.0.0.1:8080/health/ready"] interval: 5s diff --git a/evidence/api_scaling.png b/evidence/api_scaling.png index e64d327..0977f05 100644 Binary files a/evidence/api_scaling.png and b/evidence/api_scaling.png differ diff --git a/evidence/db_scaling.png b/evidence/db_scaling.png index 09ede13..b2c5b5c 100644 Binary files a/evidence/db_scaling.png and b/evidence/db_scaling.png differ diff --git a/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.json b/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.json index fc142e7..cb96c15 100644 --- a/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.json +++ b/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.json @@ -2,52 +2,73 @@ "series": "A", "api_replicas": 1, "api_cpu_limit": 1.0, + "api_mem_limit": "512m", + "api_mem_limit_bytes": 536870912, + "api_mem_total_bytes": 536870912, "db_cpu_limit": 2.0, + "http_workers": 8, "pg_pool": 8, "nginx_worker_processes": "1", + "lb_cpu_limit": 0.25, "warmup_seconds": 10, - "inspect": [ - "/mtusi2026lab3-api-1 1000000000 map[com.docker.compose.config-hash:a0d0bfc74956001c4529e42788007cb2de76144eda4a10a4ef7be0ff88f426f0 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-1 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-cache-1 500000000 map[com.docker.compose.config-hash:a7d737a55154622d18c1f35f33d3767c8335e00fb23f230f1afca917e900252d com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:32627109abf6f741121096b45c732f758876803efd7b2e1018ebc0350d117119 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:cache com.docker.compose.version:5.5.1 org.opencontainers.image.source:https://github.com/valkey-io/valkey]", - "/mtusi2026lab3-db-1 2000000000 map[com.docker.compose.config-hash:e1447a13d70e0172bc8c2b2b96bd30e491dd7eddbc4f79f3412bd6e987477e84 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:cbf42f0edc946c0bc1b8eeed62403734e1f25dc67a3491455d07a92fdb558b28 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:db-1 com.docker.compose.service:db com.docker.compose.version:5.5.1]", - "/mtusi2026lab3-lb-1 250000000 map[com.docker.compose.config-hash:8afd1a6fe04af86c6e4c4bc0c6be3e01db2e2c692a325086cc477df1a110aebb com.docker.compose.container-number:1 com.docker.compose.depends_on:api:service_healthy:false com.docker.compose.image:sha256:09ab424a8c788f8d0fe3a64429f6d19dfa526885c8609b748d0943a75dcb9f8c com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:lb com.docker.compose.version:5.5.1 maintainer:NGINX Docker Maintainers ]" - ], + "inspect": { + "verified": true, + "api": [ + { + "name": "mtusi2026lab3-api-1", + "nano_cpus": 1000000000, + "memory": 536870912, + "http_workers": "8", + "db_pool_size": "8" + } + ], + "api_memory_sum": 536870912, + "db": { + "name": "mtusi2026lab3-db-1", + "nano_cpus": 2000000000 + }, + "lb": { + "name": "mtusi2026lab3-lb-1", + "nano_cpus": 250000000, + "nginx_worker_processes": "worker_processes 1;" + } + }, "cpu": { + "lb": { + "samples": 6, + "mean_cpu_percent": 26.55, + "max_cpu_percent": 28.44 + }, "api": { "samples": 6, - "mean_cpu_percent": 90.41, - "max_cpu_percent": 100.84 + "mean_cpu_percent": 97.93, + "max_cpu_percent": 105.29 }, "db": { "samples": 6, - "mean_cpu_percent": 186.35, - "max_cpu_percent": 207.04 - }, - "lb": { - "samples": 6, - "mean_cpu_percent": 25.69, - "max_cpu_percent": 26.41 + "mean_cpu_percent": 198.84, + "max_cpu_percent": 212.46 }, "cache": { "samples": 6, - "mean_cpu_percent": 7.52, - "max_cpu_percent": 8.86 + "mean_cpu_percent": 9.22, + "max_cpu_percent": 17.24 } }, "load": { "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.758, + "elapsed_seconds": 30.705, "concurrency": 50, - "successful_requests": 10765, - "failed_requests": 0, - "error_ratio": 0.0, - "successful_rps": 349.994, + "successful_requests": 11628, + "failed_requests": 21, + "error_ratio": 0.001803, + "successful_rps": 378.703, "successful_latency_ms": { - "p50": 121.454, - "p95": 202.572, - "p99": 1115.09 + "p50": 90.288, + "p95": 204.67, + "p99": 1124.75 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.load.json b/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.load.json index 255252e..9a1c508 100644 --- a/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.load.json +++ b/evidence/track1-A-r1-apicpu1.0-dbcpu2.0.load.json @@ -2,16 +2,16 @@ "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.758, + "elapsed_seconds": 30.705, "concurrency": 50, - "successful_requests": 10765, - "failed_requests": 0, - "error_ratio": 0.0, - "successful_rps": 349.994, + "successful_requests": 11628, + "failed_requests": 21, + "error_ratio": 0.001803, + "successful_rps": 378.703, "successful_latency_ms": { - "p50": 121.454, - "p95": 202.572, - "p99": 1115.09 + "p50": 90.288, + "p95": 204.67, + "p99": 1124.75 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.json b/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.json index 3ddca7e..9e1be4a 100644 --- a/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.json +++ b/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.json @@ -2,53 +2,80 @@ "series": "A", "api_replicas": 2, "api_cpu_limit": 0.5, + "api_mem_limit": "256m", + "api_mem_limit_bytes": 268435456, + "api_mem_total_bytes": 536870912, "db_cpu_limit": 2.0, + "http_workers": 8, "pg_pool": 8, "nginx_worker_processes": "1", + "lb_cpu_limit": 0.25, "warmup_seconds": 10, - "inspect": [ - "/mtusi2026lab3-api-1 500000000 map[com.docker.compose.config-hash:9ba43bae4b038b5ebe586704e952ebccfab2ccc1d385ed1b26af3b79f702ddbd com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-1 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-2 500000000 map[com.docker.compose.config-hash:9ba43bae4b038b5ebe586704e952ebccfab2ccc1d385ed1b26af3b79f702ddbd com.docker.compose.container-number:2 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-cache-1 500000000 map[com.docker.compose.config-hash:a7d737a55154622d18c1f35f33d3767c8335e00fb23f230f1afca917e900252d com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:32627109abf6f741121096b45c732f758876803efd7b2e1018ebc0350d117119 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:cache com.docker.compose.version:5.5.1 org.opencontainers.image.source:https://github.com/valkey-io/valkey]", - "/mtusi2026lab3-db-1 2000000000 map[com.docker.compose.config-hash:e1447a13d70e0172bc8c2b2b96bd30e491dd7eddbc4f79f3412bd6e987477e84 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:cbf42f0edc946c0bc1b8eeed62403734e1f25dc67a3491455d07a92fdb558b28 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:db-1 com.docker.compose.service:db com.docker.compose.version:5.5.1]", - "/mtusi2026lab3-lb-1 250000000 map[com.docker.compose.config-hash:8afd1a6fe04af86c6e4c4bc0c6be3e01db2e2c692a325086cc477df1a110aebb com.docker.compose.container-number:1 com.docker.compose.depends_on:api:service_healthy:false com.docker.compose.image:sha256:09ab424a8c788f8d0fe3a64429f6d19dfa526885c8609b748d0943a75dcb9f8c com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:lb com.docker.compose.version:5.5.1 maintainer:NGINX Docker Maintainers ]" - ], + "inspect": { + "verified": true, + "api": [ + { + "name": "mtusi2026lab3-api-1", + "nano_cpus": 500000000, + "memory": 268435456, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-2", + "nano_cpus": 500000000, + "memory": 268435456, + "http_workers": "8", + "db_pool_size": "8" + } + ], + "api_memory_sum": 536870912, + "db": { + "name": "mtusi2026lab3-db-1", + "nano_cpus": 2000000000 + }, + "lb": { + "name": "mtusi2026lab3-lb-1", + "nano_cpus": 250000000, + "nginx_worker_processes": "worker_processes 1;" + } + }, "cpu": { + "lb": { + "samples": 6, + "mean_cpu_percent": 24.03, + "max_cpu_percent": 26.81 + }, "api": { "samples": 12, - "mean_cpu_percent": 47.38, - "max_cpu_percent": 55.02 + "mean_cpu_percent": 43.88, + "max_cpu_percent": 53.96 }, "db": { "samples": 6, - "mean_cpu_percent": 166.57, - "max_cpu_percent": 204.93 - }, - "lb": { - "samples": 6, - "mean_cpu_percent": 19.7, - "max_cpu_percent": 24.94 + "mean_cpu_percent": 166.58, + "max_cpu_percent": 192.08 }, "cache": { "samples": 6, - "mean_cpu_percent": 8.12, - "max_cpu_percent": 15.09 + "mean_cpu_percent": 7.98, + "max_cpu_percent": 16.26 } }, "load": { "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.161, + "elapsed_seconds": 30.231, "concurrency": 50, - "successful_requests": 9881, + "successful_requests": 10463, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 327.611, + "successful_rps": 346.098, "successful_latency_ms": { - "p50": 131.433, - "p95": 289.233, - "p99": 404.102 + "p50": 128.325, + "p95": 269.126, + "p99": 380.157 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.load.json b/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.load.json index a68b429..a9c7c7a 100644 --- a/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.load.json +++ b/evidence/track1-A-r2-apicpu0.5-dbcpu2.0.load.json @@ -2,16 +2,16 @@ "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.161, + "elapsed_seconds": 30.231, "concurrency": 50, - "successful_requests": 9881, + "successful_requests": 10463, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 327.611, + "successful_rps": 346.098, "successful_latency_ms": { - "p50": 131.433, - "p95": 289.233, - "p99": 404.102 + "p50": 128.325, + "p95": 269.126, + "p99": 380.157 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.json b/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.json index cc4d42e..eac6533 100644 --- a/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.json +++ b/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.json @@ -2,55 +2,94 @@ "series": "A", "api_replicas": 4, "api_cpu_limit": 0.25, + "api_mem_limit": "128m", + "api_mem_limit_bytes": 134217728, + "api_mem_total_bytes": 536870912, "db_cpu_limit": 2.0, + "http_workers": 8, "pg_pool": 8, "nginx_worker_processes": "1", + "lb_cpu_limit": 0.25, "warmup_seconds": 10, - "inspect": [ - "/mtusi2026lab3-api-1 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-1 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-2 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:2 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-2 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-3 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:3 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-4 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:4 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-cache-1 500000000 map[com.docker.compose.config-hash:a7d737a55154622d18c1f35f33d3767c8335e00fb23f230f1afca917e900252d com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:32627109abf6f741121096b45c732f758876803efd7b2e1018ebc0350d117119 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:cache com.docker.compose.version:5.5.1 org.opencontainers.image.source:https://github.com/valkey-io/valkey]", - "/mtusi2026lab3-db-1 2000000000 map[com.docker.compose.config-hash:e1447a13d70e0172bc8c2b2b96bd30e491dd7eddbc4f79f3412bd6e987477e84 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:cbf42f0edc946c0bc1b8eeed62403734e1f25dc67a3491455d07a92fdb558b28 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:db-1 com.docker.compose.service:db com.docker.compose.version:5.5.1]", - "/mtusi2026lab3-lb-1 250000000 map[com.docker.compose.config-hash:8afd1a6fe04af86c6e4c4bc0c6be3e01db2e2c692a325086cc477df1a110aebb com.docker.compose.container-number:1 com.docker.compose.depends_on:api:service_healthy:false com.docker.compose.image:sha256:09ab424a8c788f8d0fe3a64429f6d19dfa526885c8609b748d0943a75dcb9f8c com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:lb com.docker.compose.version:5.5.1 maintainer:NGINX Docker Maintainers ]" - ], + "inspect": { + "verified": true, + "api": [ + { + "name": "mtusi2026lab3-api-1", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-2", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-3", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-4", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + } + ], + "api_memory_sum": 536870912, + "db": { + "name": "mtusi2026lab3-db-1", + "nano_cpus": 2000000000 + }, + "lb": { + "name": "mtusi2026lab3-lb-1", + "nano_cpus": 250000000, + "nginx_worker_processes": "worker_processes 1;" + } + }, "cpu": { + "lb": { + "samples": 7, + "mean_cpu_percent": 17.02, + "max_cpu_percent": 25.95 + }, "api": { "samples": 28, - "mean_cpu_percent": 18.64, - "max_cpu_percent": 27.51 + "mean_cpu_percent": 20.06, + "max_cpu_percent": 27.19 }, "db": { "samples": 7, - "mean_cpu_percent": 124.05, - "max_cpu_percent": 189.16 - }, - "lb": { - "samples": 7, - "mean_cpu_percent": 19.08, - "max_cpu_percent": 26.16 + "mean_cpu_percent": 139.19, + "max_cpu_percent": 210.12 }, "cache": { "samples": 7, - "mean_cpu_percent": 9.49, - "max_cpu_percent": 14.99 + "mean_cpu_percent": 10.01, + "max_cpu_percent": 17.01 } }, "load": { "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.347, + "elapsed_seconds": 30.276, "concurrency": 50, - "successful_requests": 9711, + "successful_requests": 9915, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 320.002, + "successful_rps": 327.486, "successful_latency_ms": { - "p50": 151.294, - "p95": 295.941, - "p99": 407.03 + "p50": 102.151, + "p95": 487.734, + "p99": 626.914 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.load.json b/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.load.json index c4e6310..afb70d9 100644 --- a/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.load.json +++ b/evidence/track1-A-r4-apicpu0.25-dbcpu2.0.load.json @@ -2,16 +2,16 @@ "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.347, + "elapsed_seconds": 30.276, "concurrency": 50, - "successful_requests": 9711, + "successful_requests": 9915, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 320.002, + "successful_rps": 327.486, "successful_latency_ms": { - "p50": 151.294, - "p95": 295.941, - "p99": 407.03 + "p50": 102.151, + "p95": 487.734, + "p99": 626.914 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.json b/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.json index bf5293e..ecb2e1b 100644 --- a/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.json +++ b/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.json @@ -2,55 +2,94 @@ "series": "B", "api_replicas": 4, "api_cpu_limit": 0.25, + "api_mem_limit": "128m", + "api_mem_limit_bytes": 134217728, + "api_mem_total_bytes": 536870912, "db_cpu_limit": 1.0, + "http_workers": 8, "pg_pool": 8, "nginx_worker_processes": "1", + "lb_cpu_limit": 0.25, "warmup_seconds": 10, - "inspect": [ - "/mtusi2026lab3-api-1 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-1 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-2 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:2 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-2 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-3 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:3 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-3 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-4 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:4 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-4 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-cache-1 500000000 map[com.docker.compose.config-hash:a7d737a55154622d18c1f35f33d3767c8335e00fb23f230f1afca917e900252d com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:32627109abf6f741121096b45c732f758876803efd7b2e1018ebc0350d117119 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:cache com.docker.compose.version:5.5.1 org.opencontainers.image.source:https://github.com/valkey-io/valkey]", - "/mtusi2026lab3-db-1 1000000000 map[com.docker.compose.config-hash:5a9910bcb7033eb32e9c8f69ff91ca585ca4927069203321b91d33cdf049f2d2 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:cbf42f0edc946c0bc1b8eeed62403734e1f25dc67a3491455d07a92fdb558b28 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:db-1 com.docker.compose.service:db com.docker.compose.version:5.5.1]", - "/mtusi2026lab3-lb-1 250000000 map[com.docker.compose.config-hash:8afd1a6fe04af86c6e4c4bc0c6be3e01db2e2c692a325086cc477df1a110aebb com.docker.compose.container-number:1 com.docker.compose.depends_on:api:service_healthy:false com.docker.compose.image:sha256:09ab424a8c788f8d0fe3a64429f6d19dfa526885c8609b748d0943a75dcb9f8c com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:lb com.docker.compose.version:5.5.1 maintainer:NGINX Docker Maintainers ]" - ], + "inspect": { + "verified": true, + "api": [ + { + "name": "mtusi2026lab3-api-1", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-2", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-3", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-4", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + } + ], + "api_memory_sum": 536870912, + "db": { + "name": "mtusi2026lab3-db-1", + "nano_cpus": 1000000000 + }, + "lb": { + "name": "mtusi2026lab3-lb-1", + "nano_cpus": 250000000, + "nginx_worker_processes": "worker_processes 1;" + } + }, "cpu": { + "lb": { + "samples": 7, + "mean_cpu_percent": 8.62, + "max_cpu_percent": 11.8 + }, "api": { "samples": 28, - "mean_cpu_percent": 9.62, - "max_cpu_percent": 14.31 + "mean_cpu_percent": 11.86, + "max_cpu_percent": 19.43 }, "db": { "samples": 7, - "mean_cpu_percent": 73.45, - "max_cpu_percent": 107.26 - }, - "lb": { - "samples": 7, - "mean_cpu_percent": 7.77, - "max_cpu_percent": 11.61 + "mean_cpu_percent": 75.17, + "max_cpu_percent": 102.55 }, "cache": { "samples": 7, - "mean_cpu_percent": 6.37, - "max_cpu_percent": 10.33 + "mean_cpu_percent": 6.74, + "max_cpu_percent": 11.3 } }, "load": { "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.313, + "elapsed_seconds": 30.333, "concurrency": 50, - "successful_requests": 5021, + "successful_requests": 5428, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 165.641, + "successful_rps": 178.949, "successful_latency_ms": { - "p50": 297.372, - "p95": 499.711, - "p99": 598.3 + "p50": 286.653, + "p95": 480.072, + "p99": 538.062 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.load.json b/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.load.json index 0718116..d596aed 100644 --- a/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.load.json +++ b/evidence/track1-B-r4-apicpu0.25-dbcpu1.0.load.json @@ -2,16 +2,16 @@ "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.313, + "elapsed_seconds": 30.333, "concurrency": 50, - "successful_requests": 5021, + "successful_requests": 5428, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 165.641, + "successful_rps": 178.949, "successful_latency_ms": { - "p50": 297.372, - "p95": 499.711, - "p99": 598.3 + "p50": 286.653, + "p95": 480.072, + "p99": 538.062 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.json b/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.json index 751a39d..33815e4 100644 --- a/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.json +++ b/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.json @@ -2,55 +2,94 @@ "series": "B", "api_replicas": 4, "api_cpu_limit": 0.25, + "api_mem_limit": "128m", + "api_mem_limit_bytes": 134217728, + "api_mem_total_bytes": 536870912, "db_cpu_limit": 2.0, + "http_workers": 8, "pg_pool": 8, "nginx_worker_processes": "1", + "lb_cpu_limit": 0.25, "warmup_seconds": 10, - "inspect": [ - "/mtusi2026lab3-api-1 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-1 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-2 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:2 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-2 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-3 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:3 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-3 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-4 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:4 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-4 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-cache-1 500000000 map[com.docker.compose.config-hash:a7d737a55154622d18c1f35f33d3767c8335e00fb23f230f1afca917e900252d com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:32627109abf6f741121096b45c732f758876803efd7b2e1018ebc0350d117119 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:cache com.docker.compose.version:5.5.1 org.opencontainers.image.source:https://github.com/valkey-io/valkey]", - "/mtusi2026lab3-db-1 2000000000 map[com.docker.compose.config-hash:e1447a13d70e0172bc8c2b2b96bd30e491dd7eddbc4f79f3412bd6e987477e84 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:cbf42f0edc946c0bc1b8eeed62403734e1f25dc67a3491455d07a92fdb558b28 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:db-1 com.docker.compose.service:db com.docker.compose.version:5.5.1]", - "/mtusi2026lab3-lb-1 250000000 map[com.docker.compose.config-hash:8afd1a6fe04af86c6e4c4bc0c6be3e01db2e2c692a325086cc477df1a110aebb com.docker.compose.container-number:1 com.docker.compose.depends_on:api:service_healthy:false com.docker.compose.image:sha256:09ab424a8c788f8d0fe3a64429f6d19dfa526885c8609b748d0943a75dcb9f8c com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:lb com.docker.compose.version:5.5.1 maintainer:NGINX Docker Maintainers ]" - ], + "inspect": { + "verified": true, + "api": [ + { + "name": "mtusi2026lab3-api-1", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-2", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-3", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-4", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + } + ], + "api_memory_sum": 536870912, + "db": { + "name": "mtusi2026lab3-db-1", + "nano_cpus": 2000000000 + }, + "lb": { + "name": "mtusi2026lab3-lb-1", + "nano_cpus": 250000000, + "nginx_worker_processes": "worker_processes 1;" + } + }, "cpu": { + "lb": { + "samples": 7, + "mean_cpu_percent": 16.89, + "max_cpu_percent": 25.75 + }, "api": { "samples": 28, - "mean_cpu_percent": 19.77, - "max_cpu_percent": 27.63 + "mean_cpu_percent": 20.09, + "max_cpu_percent": 28.18 }, "db": { "samples": 7, - "mean_cpu_percent": 136.68, - "max_cpu_percent": 206.58 - }, - "lb": { - "samples": 7, - "mean_cpu_percent": 16.56, - "max_cpu_percent": 25.95 + "mean_cpu_percent": 131.47, + "max_cpu_percent": 190.53 }, "cache": { "samples": 7, - "mean_cpu_percent": 5.61, - "max_cpu_percent": 7.95 + "mean_cpu_percent": 9.64, + "max_cpu_percent": 19.24 } }, "load": { "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.391, + "elapsed_seconds": 30.491, "concurrency": 50, - "successful_requests": 10082, + "successful_requests": 10191, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 331.742, + "successful_rps": 334.227, "successful_latency_ms": { - "p50": 112.739, - "p95": 377.522, - "p99": 459.163 + "p50": 117.124, + "p95": 366.356, + "p99": 455.007 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.load.json b/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.load.json index ac0ef77..d6946a1 100644 --- a/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.load.json +++ b/evidence/track1-B-r4-apicpu0.25-dbcpu2.0.load.json @@ -2,16 +2,16 @@ "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.391, + "elapsed_seconds": 30.491, "concurrency": 50, - "successful_requests": 10082, + "successful_requests": 10191, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 331.742, + "successful_rps": 334.227, "successful_latency_ms": { - "p50": 112.739, - "p95": 377.522, - "p99": 459.163 + "p50": 117.124, + "p95": 366.356, + "p99": 455.007 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.json b/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.json index 226f15a..40d3b83 100644 --- a/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.json +++ b/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.json @@ -2,55 +2,94 @@ "series": "B", "api_replicas": 4, "api_cpu_limit": 0.25, + "api_mem_limit": "128m", + "api_mem_limit_bytes": 134217728, + "api_mem_total_bytes": 536870912, "db_cpu_limit": 4.0, + "http_workers": 8, "pg_pool": 8, "nginx_worker_processes": "1", + "lb_cpu_limit": 0.25, "warmup_seconds": 10, - "inspect": [ - "/mtusi2026lab3-api-1 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-1 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-2 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:2 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-2 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-3 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:3 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-3 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-api-4 250000000 map[com.docker.compose.config-hash:baa6c518a5d0d2ce1a2d38ddcfa206234dfa8023729d82b99b3f7264efaf99b4 com.docker.compose.container-number:4 com.docker.compose.depends_on: com.docker.compose.image:sha256:a1ec572a54fa12c0ab467a444062c9a59e7bc804b20247d743295cee651b6fdc com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:api-4 com.docker.compose.service:api com.docker.compose.version:5.5.1 org.opencontainers.image.version:25.10]", - "/mtusi2026lab3-cache-1 500000000 map[com.docker.compose.config-hash:a7d737a55154622d18c1f35f33d3767c8335e00fb23f230f1afca917e900252d com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:32627109abf6f741121096b45c732f758876803efd7b2e1018ebc0350d117119 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:cache com.docker.compose.version:5.5.1 org.opencontainers.image.source:https://github.com/valkey-io/valkey]", - "/mtusi2026lab3-db-1 4000000000 map[com.docker.compose.config-hash:00262326412a2d2388e383afd1f1c5d3453baeebf2e459788f9a579b8340f994 com.docker.compose.container-number:1 com.docker.compose.depends_on: com.docker.compose.image:sha256:cbf42f0edc946c0bc1b8eeed62403734e1f25dc67a3491455d07a92fdb558b28 com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.replace:db-1 com.docker.compose.service:db com.docker.compose.version:5.5.1]", - "/mtusi2026lab3-lb-1 250000000 map[com.docker.compose.config-hash:8afd1a6fe04af86c6e4c4bc0c6be3e01db2e2c692a325086cc477df1a110aebb com.docker.compose.container-number:1 com.docker.compose.depends_on:api:service_healthy:false com.docker.compose.image:sha256:09ab424a8c788f8d0fe3a64429f6d19dfa526885c8609b748d0943a75dcb9f8c com.docker.compose.oneoff:False com.docker.compose.project:mtusi2026lab3 com.docker.compose.project.config_files:/home/andrey/mtusi2026lab3/compose.yaml com.docker.compose.project.environment_file:/home/andrey/mtusi2026lab3/.env com.docker.compose.project.working_dir:/home/andrey/mtusi2026lab3 com.docker.compose.service:lb com.docker.compose.version:5.5.1 maintainer:NGINX Docker Maintainers ]" - ], + "inspect": { + "verified": true, + "api": [ + { + "name": "mtusi2026lab3-api-1", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-2", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-3", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + }, + { + "name": "mtusi2026lab3-api-4", + "nano_cpus": 250000000, + "memory": 134217728, + "http_workers": "8", + "db_pool_size": "8" + } + ], + "api_memory_sum": 536870912, + "db": { + "name": "mtusi2026lab3-db-1", + "nano_cpus": 4000000000 + }, + "lb": { + "name": "mtusi2026lab3-lb-1", + "nano_cpus": 250000000, + "nginx_worker_processes": "worker_processes 1;" + } + }, "cpu": { + "lb": { + "samples": 7, + "mean_cpu_percent": 17.63, + "max_cpu_percent": 25.98 + }, "api": { "samples": 28, - "mean_cpu_percent": 19.49, - "max_cpu_percent": 27.41 + "mean_cpu_percent": 19.88, + "max_cpu_percent": 27.94 }, "db": { "samples": 7, - "mean_cpu_percent": 141.25, - "max_cpu_percent": 241.9 - }, - "lb": { - "samples": 7, - "mean_cpu_percent": 16.49, - "max_cpu_percent": 26.16 + "mean_cpu_percent": 138.77, + "max_cpu_percent": 203.66 }, "cache": { "samples": 7, - "mean_cpu_percent": 6.55, - "max_cpu_percent": 8.51 + "mean_cpu_percent": 10.56, + "max_cpu_percent": 17.54 } }, "load": { "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.223, + "elapsed_seconds": 30.31, "concurrency": 50, - "successful_requests": 12619, + "successful_requests": 10578, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 417.528, + "successful_rps": 348.996, "successful_latency_ms": { - "p50": 99.961, - "p95": 288.263, - "p99": 371.702 + "p50": 110.449, + "p95": 394.75, + "p99": 479.704 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.load.json b/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.load.json index 821c5a9..e7f1a60 100644 --- a/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.load.json +++ b/evidence/track1-B-r4-apicpu0.25-dbcpu4.0.load.json @@ -2,16 +2,16 @@ "lab": 3, "workload": "closed-loop POST messages; no automatic retries", "requested_duration_seconds": 30.0, - "elapsed_seconds": 30.223, + "elapsed_seconds": 30.31, "concurrency": 50, - "successful_requests": 12619, + "successful_requests": 10578, "failed_requests": 0, "error_ratio": 0.0, - "successful_rps": 417.528, + "successful_rps": 348.996, "successful_latency_ms": { - "p50": 99.961, - "p95": 288.263, - "p99": 371.702 + "p50": 110.449, + "p95": 394.75, + "p99": 479.704 }, "note": "Latency percentiles cover successes only. Timeouts/errors are counted separately; closed-loop has coordinated omission and client overhead." } diff --git a/lb/nginx.conf b/lb/nginx.conf.template similarity index 93% rename from lb/nginx.conf rename to lb/nginx.conf.template index db5a272..a484360 100644 --- a/lb/nginx.conf +++ b/lb/nginx.conf.template @@ -1,9 +1,9 @@ -worker_processes auto; +worker_processes ${NGINX_WORKERS}; error_log /dev/stderr warn; pid /tmp/nginx.pid; events { - worker_connections 4096; + worker_connections 1024; } http { diff --git a/scripts/plot_graphs.py b/scripts/plot_graphs.py index a0283fc..aaca19d 100644 --- a/scripts/plot_graphs.py +++ b/scripts/plot_graphs.py @@ -55,8 +55,6 @@ def dual_axis(ax, xs, rps, p99, xlabel, title): def main(): series_a = sorted(load_series("A"), key=lambda row: row["replicas"]) series_b = sorted(load_series("B"), key=lambda row: row["db_cpu"]) - repeat = load_series("R") - tuned = load_series("P") if not series_a or not series_b: raise SystemExit("need series A and B JSON in evidence/") @@ -66,30 +64,22 @@ def main(): [row["replicas"] for row in series_a], [row["rps"] for row in series_a], [row["p99"] for row in series_a], - "API replicas (total CPU = 1.0, DB CPU = 2.0)", - "Серия A: реплики API при фиксированном суммарном CPU", + "API replicas (total CPU = 1.0, total RAM = 512 MiB, DB CPU = 2.0)", + "Серия A: реплики API при фиксированных суммарных CPU и RAM", ) fig.tight_layout() fig.savefig(EVIDENCE / "api_scaling.png", dpi=140) plt.close(fig) fig, ax = plt.subplots(figsize=(7.2, 4.4)) - twin = dual_axis( + dual_axis( ax, [row["db_cpu"] for row in series_b], [row["rps"] for row in series_b], [row["p99"] for row in series_b], - "PostgreSQL CPU limit (4 API × 0.25 CPU)", - "Серия B: CPU базы и парный прогон тюнинга", + "PostgreSQL CPU limit (4 API × 0.25 CPU, total API RAM = 512 MiB)", + "Серия B: RPS и p99 от CPU базы", ) - if repeat and tuned: - base, point = repeat[-1], tuned[-1] - ax.scatter([base["db_cpu"]], [base["rps"]], marker="x", s=80, color="#1f4e79", zorder=5, label="repeat baseline RPS") - ax.scatter([point["db_cpu"] + 0.08], [point["rps"]], marker="*", s=160, color="#1f4e79", zorder=5, label="tuned RPS") - twin.scatter([base["db_cpu"]], [base["p99"]], marker="x", s=80, color="#b85c38", zorder=5, label="repeat baseline p99") - twin.scatter([point["db_cpu"] + 0.08], [point["p99"]], marker="*", s=160, color="#b85c38", zorder=5, label="tuned p99") - handles = ax.get_lines() + twin.get_lines() + list(ax.collections) + list(twin.collections) - ax.legend(handles, [h.get_label() for h in handles], loc="best", fontsize=8) fig.tight_layout() fig.savefig(EVIDENCE / "db_scaling.png", dpi=140) plt.close(fig) diff --git a/scripts/run_track1.py b/scripts/run_track1.py index bc5bc88..d789acf 100755 --- a/scripts/run_track1.py +++ b/scripts/run_track1.py @@ -1,5 +1,9 @@ #!/usr/bin/env python3 -"""One measured load.py run under explicit Compose CPU limits. Prints nothing secret.""" +"""Series A and B with a fixed total API RAM budget and the base 8/8/1 profile. + +The profile is applied through Compose environment, then checked with docker inspect +before the load window. Nothing secret is printed. +""" import json import os import subprocess @@ -8,30 +12,66 @@ import threading import time import urllib.error import urllib.request +from decimal import Decimal from pathlib import Path ROOT = Path(__file__).resolve().parents[1] +HTTP_WORKERS = "8" +DB_POOL_SIZE = "8" +NGINX_WORKERS = "1" +LB_CPU_LIMIT = "0.25" +TOTAL_API_RAM = 512 * 1024 * 1024 -def env_for(api_replicas, api_cpu, db_cpu): +def mem_bytes(spec): + text = str(spec).strip() + units = {"b": 1, "k": 1024, "m": 1024**2, "g": 1024**3} + unit = text[-1].lower() + if unit not in units: + raise SystemExit(f"bad memory spec: {spec}") + return int(text[:-1]) * units[unit] + + +def cpu_nanos(spec): + return int(Decimal(str(spec)) * Decimal(1_000_000_000)) + + +def series_points(): + points = [] + for replicas, api_cpu in ((1, "1.0"), (2, "0.5"), (4, "0.25")): + per = TOTAL_API_RAM // replicas // (1024 * 1024) + points.append(("A", replicas, api_cpu, f"{per}m", "2.0")) + for db_cpu in ("1.0", "2.0", "4.0"): + per = TOTAL_API_RAM // 4 // (1024 * 1024) + points.append(("B", 4, "0.25", f"{per}m", db_cpu)) + return points + + +def env_for(replicas, api_cpu, api_mem, db_cpu): env = os.environ.copy() - env["API_REPLICAS"] = str(api_replicas) + env["API_REPLICAS"] = str(replicas) env["API_CPU_LIMIT"] = str(api_cpu) + env["API_MEM_LIMIT"] = api_mem env["DB_CPU_LIMIT"] = str(db_cpu) + env["HTTP_WORKERS"] = HTTP_WORKERS + env["DB_POOL_SIZE"] = DB_POOL_SIZE + env["NGINX_WORKERS"] = NGINX_WORKERS + env["LB_CPU_LIMIT"] = LB_CPU_LIMIT return env -def compose(args, env): +def compose(args, env, capture=False): result = subprocess.run( ["docker", "compose", "--env-file", str(ROOT / ".env"), *args], cwd=ROOT, env=env, check=False, text=True, - capture_output=True, + capture_output=capture, ) - if result.returncode != 0: - sys.stderr.write(result.stderr or result.stdout) + if result.returncode != 0 and capture: + sys.stderr.write(result.stderr or "") + sys.stderr.write(result.stdout or "") return result @@ -39,7 +79,7 @@ def wait_stack(env, replicas, timeout=180): deadline = time.monotonic() + timeout last = "" while time.monotonic() < deadline: - ps = compose(["ps", "api", "--format", "{{.Health}}"], env) + ps = compose(["ps", "api", "--format", "{{.Health}}"], env, capture=True) health = [line.strip() for line in ps.stdout.splitlines() if line.strip()] ready = False try: @@ -54,18 +94,118 @@ def wait_stack(env, replicas, timeout=180): raise SystemExit("stack did not become ready: " + last) -def observed_nanos(env): - ids = compose(["ps", "-q"], env).stdout.split() - found = [] - for cid in ids: - info = subprocess.check_output( - ["docker", "inspect", "-f", "{{.Name}} {{.HostConfig.NanoCpus}} {{.Config.Labels}}", cid], - text=True, - ).strip() - found.append(info) +def container_ids(env, service): + result = compose(["ps", "-q", service], env, capture=True) + if result.returncode != 0: + raise SystemExit(f"compose ps {service} failed") + return [line for line in result.stdout.split() if line] + + +def inspect_container(cid): + raw = subprocess.check_output(["docker", "inspect", cid], text=True) + return json.loads(raw)[0] + + +def env_map(info): + found = {} + for item in info["Config"].get("Env") or []: + key, _, value = item.partition("=") + found[key] = value return found +def verify_limits(env, replicas, api_cpu, api_mem, db_cpu): + expect_api_nanos = cpu_nanos(api_cpu) + expect_api_mem = mem_bytes(api_mem) + expect_db_nanos = cpu_nanos(db_cpu) + expect_lb_nanos = cpu_nanos(LB_CPU_LIMIT) + api_ids = container_ids(env, "api") + if len(api_ids) != replicas: + raise SystemExit(f"expected {replicas} api containers, found {len(api_ids)}") + + api_rows = [] + for cid in api_ids: + info = inspect_container(cid) + nanos = info["HostConfig"]["NanoCpus"] + memory = info["HostConfig"]["Memory"] + present = env_map(info) + name = info["Name"].lstrip("/") + if nanos != expect_api_nanos or memory != expect_api_mem: + raise SystemExit( + f"{name}: NanoCpus={nanos} Memory={memory}, " + f"expected {expect_api_nanos} and {expect_api_mem}" + ) + if present.get("HTTP_WORKERS") != HTTP_WORKERS or present.get("DB_POOL_SIZE") != DB_POOL_SIZE: + raise SystemExit( + f"{name}: HTTP_WORKERS={present.get('HTTP_WORKERS')} " + f"DB_POOL_SIZE={present.get('DB_POOL_SIZE')}" + ) + api_rows.append( + { + "name": name, + "nano_cpus": nanos, + "memory": memory, + "http_workers": present["HTTP_WORKERS"], + "db_pool_size": present["DB_POOL_SIZE"], + } + ) + total_mem = sum(row["memory"] for row in api_rows) + if total_mem != TOTAL_API_RAM: + raise SystemExit(f"api RAM sum {total_mem} != {TOTAL_API_RAM}") + + db_ids = container_ids(env, "db") + if len(db_ids) != 1: + raise SystemExit(f"expected 1 db container, found {len(db_ids)}") + db_info = inspect_container(db_ids[0]) + db_nanos = db_info["HostConfig"]["NanoCpus"] + if db_nanos != expect_db_nanos: + raise SystemExit(f"db NanoCpus={db_nanos}, expected {expect_db_nanos}") + + lb_ids = container_ids(env, "lb") + if len(lb_ids) != 1: + raise SystemExit(f"expected 1 lb container, found {len(lb_ids)}") + lb_info = inspect_container(lb_ids[0]) + lb_nanos = lb_info["HostConfig"]["NanoCpus"] + lb_env = env_map(lb_info) + if lb_nanos != expect_lb_nanos or lb_env.get("NGINX_WORKERS") != NGINX_WORKERS: + raise SystemExit( + f"lb NanoCpus={lb_nanos} NGINX_WORKERS={lb_env.get('NGINX_WORKERS')}" + ) + rendered = compose(["exec", "-T", "lb", "nginx", "-T"], env, capture=True) + if rendered.returncode != 0: + raise SystemExit("nginx -T failed") + directive = "" + rendered_text = (rendered.stdout or "") + "\n" + (rendered.stderr or "") + for line in rendered_text.splitlines(): + stripped = line.strip() + if stripped.startswith("worker_processes"): + directive = stripped + break + expected_directive = f"worker_processes {NGINX_WORKERS};" + if directive != expected_directive: + raise SystemExit(f"nginx config has {directive!r}, expected {expected_directive!r}") + + logs = compose(["logs", "api"], env, capture=True) + if logs.returncode != 0: + raise SystemExit("api logs failed") + pool_line = f"pg pool ready ({DB_POOL_SIZE} connections)" + worker_line = f"http workers: {HTTP_WORKERS}" + if logs.stdout.count(pool_line) < replicas or logs.stdout.count(worker_line) < replicas: + raise SystemExit("api logs do not show the base pool and worker profile on every replica") + + return { + "verified": True, + "api": api_rows, + "api_memory_sum": total_mem, + "db": {"name": db_info["Name"].lstrip("/"), "nano_cpus": db_nanos}, + "lb": { + "name": lb_info["Name"].lstrip("/"), + "nano_cpus": lb_nanos, + "nginx_worker_processes": directive, + }, + } + + def sample_stats(stop, bucket): while not stop.is_set(): proc = subprocess.run( @@ -81,9 +221,8 @@ def sample_stats(stop, bucket): if len(parts) != 3: continue name, cpu, mem = parts - cpu_val = cpu.strip().rstrip("%") try: - cpu_num = float(cpu_val) + cpu_num = float(cpu.strip().rstrip("%")) except ValueError: continue rows.append({"name": name, "cpu_percent": cpu_num, "mem": mem.strip()}) @@ -113,43 +252,34 @@ def summarize_cpu(samples): return out -def main(): - if len(sys.argv) != 7: - raise SystemExit( - "usage: run_track1.py SERIES REPLICAS API_CPU DB_CPU POOL NGINX_WORKERS" - ) - series, replicas, api_cpu, db_cpu, pool, nginx_workers = sys.argv[1:] - replicas = int(replicas) - env = env_for(replicas, api_cpu, db_cpu) +def run_point(series, replicas, api_cpu, api_mem, db_cpu): + env = env_for(replicas, api_cpu, api_mem, db_cpu) label = f"track1-{series}-r{replicas}-apicpu{api_cpu}-dbcpu{db_cpu}" raw = ROOT / "evidence" / f"{label}.load.json" dest = ROOT / "evidence" / f"{label}.json" - - up_db = None - for _ in range(3): - up_db = compose( - ["up", "-d", "--no-deps", "--force-recreate", "db"], - env, - ) - if up_db.returncode == 0: - break - time.sleep(2) - if up_db.returncode != 0: - sys.stderr.write(up_db.stderr) - raise SystemExit(up_db.returncode) - up_api = compose( - ["up", "-d", "--no-deps", "--force-recreate", "--scale", f"api={replicas}", "api"], + print( + f"point series={series} replicas={replicas} api_cpu={api_cpu} " + f"api_mem={api_mem} db_cpu={db_cpu} http_workers={HTTP_WORKERS} " + f"db_pool={DB_POOL_SIZE} nginx_workers={NGINX_WORKERS}", + file=sys.stderr, + ) + down = compose(["down"], env) + if down.returncode != 0: + raise SystemExit(down.returncode) + up = compose( + ["up", "-d", "--build", "--scale", f"api={replicas}"], env, ) - if up_api.returncode != 0: - sys.stderr.write(up_api.stderr) - raise SystemExit(up_api.returncode) - up_lb = compose(["up", "-d", "--no-deps", "lb"], env) - if up_lb.returncode != 0: - sys.stderr.write(up_lb.stderr) - raise SystemExit(up_lb.returncode) + if up.returncode != 0: + raise SystemExit(up.returncode) wait_stack(env, replicas) - limits = observed_nanos(env) + limits = verify_limits(env, replicas, api_cpu, api_mem, db_cpu) + print( + f"inspect ok api_sum_memory={limits['api_memory_sum']} " + f"api_nano={limits['api'][0]['nano_cpus']} db_nano={limits['db']['nano_cpus']} " + f"nginx={limits['lb']['nginx_worker_processes']}", + file=sys.stderr, + ) warmup = subprocess.run( [ @@ -203,9 +333,14 @@ def main(): "series": series, "api_replicas": replicas, "api_cpu_limit": float(api_cpu), + "api_mem_limit": api_mem, + "api_mem_limit_bytes": mem_bytes(api_mem), + "api_mem_total_bytes": TOTAL_API_RAM, "db_cpu_limit": float(db_cpu), - "pg_pool": int(pool), - "nginx_worker_processes": nginx_workers, + "http_workers": int(HTTP_WORKERS), + "pg_pool": int(DB_POOL_SIZE), + "nginx_worker_processes": NGINX_WORKERS, + "lb_cpu_limit": float(LB_CPU_LIMIT), "warmup_seconds": 10, "inspect": limits, "cpu": summarize_cpu(samples), @@ -213,21 +348,26 @@ def main(): } dest.write_text(json.dumps(report, ensure_ascii=False, indent=2) + "\n") lat = load["successful_latency_ms"] - print( - json.dumps( - { - "file": str(dest.relative_to(ROOT)), - "rps": load["successful_rps"], - "p50": lat["p50"], - "p95": lat["p95"], - "p99": lat["p99"], - "errors": load["failed_requests"], - "error_ratio": load["error_ratio"], - "cpu": report["cpu"], - }, - ensure_ascii=False, - ) - ) + summary = { + "file": str(dest.relative_to(ROOT)), + "rps": load["successful_rps"], + "p50": lat["p50"], + "p95": lat["p95"], + "p99": lat["p99"], + "errors": load["failed_requests"], + "error_ratio": load["error_ratio"], + "api_mem_total_bytes": TOTAL_API_RAM, + "cpu": report["cpu"], + } + print(json.dumps(summary, ensure_ascii=False)) + return summary + + +def main(): + if len(sys.argv) != 1: + raise SystemExit("usage: python3 scripts/run_track1.py") + for series, replicas, api_cpu, api_mem, db_cpu in series_points(): + run_point(series, replicas, api_cpu, api_mem, db_cpu) if __name__ == "__main__": diff --git a/src/main.cpp b/src/main.cpp index 0ad0a2b..c35f6ea 100644 --- a/src/main.cpp +++ b/src/main.cpp @@ -1192,6 +1192,20 @@ static void handle_client(int client, ::close(client); } +int env_or_default(const char* name, int fallback) { + const char* raw = std::getenv(name); + if (raw == nullptr || raw[0] == '\0') { + return fallback; + } + char* end = nullptr; + errno = 0; + const long value = std::strtol(raw, &end, 10); + if (end == raw || *end != '\0' || errno != 0 || value < 1 || value > 4096) { + return fallback; + } + return static_cast(value); +} + int main() { const char* host_env = std::getenv("APP_HOST"); const char* port_env = std::getenv("APP_PORT"); @@ -1218,8 +1232,9 @@ int main() { std::clog << "migrations ok\n"; } - PgPool pool(db_url, 32); - std::clog << "pg pool ready (32 connections)\n"; + const int pool_n = env_or_default("DB_POOL_SIZE", 8); + PgPool pool(db_url, pool_n); + std::clog << "pg pool ready (" << pool_n << " connections)\n"; UserRepo users(pool); ChatRepo chats(pool); @@ -1284,7 +1299,8 @@ int main() { std::mutex q_mu; std::condition_variable q_cv; std::queue q; - const int workers_n = 32; + const int workers_n = env_or_default("HTTP_WORKERS", 8); + std::clog << "http workers: " << workers_n << "\n"; std::vector workers; for (int i = 0; i < workers_n; ++i) { workers.emplace_back([&]() {