forked from mtusi2026/mtusi2026lab6
742 lines
28 KiB
YAML
742 lines
28 KiB
YAML
# Лабораторная 6: стенд lab5 + наблюдаемость.
|
||
# API x N / worker --OTLP/HTTP--> otel-collector --> Prometheus (метрики, alert rules)
|
||
# --> Loki (логи)
|
||
# --> Tempo (трассы)
|
||
# Grafana (127.0.0.1:3000) читает все три; provisioning и dashboard — infra/grafana.
|
||
# Сеть telemetry: api, worker, collector, хранилища, grafana. Недоступность любого
|
||
# telemetry-сервиса не блокирует API (нет depends_on, экспорт асинхронный и ограничен).
|
||
#
|
||
# База lab5: HAProxy -> API x N (по умолчанию 2) -> PostgreSQL (метаданные, задания, outbox)
|
||
# + Redis (кэш сессий)
|
||
# + SeaweedFS S3 (staging/ и final/ в приватном бакете)
|
||
# API/worker outbox relay -> broker (Redis Streams, AOF fsync always) -> worker x M
|
||
# Сети: backend — lb, api, worker, postgres, redis; storage (internal) — api, worker, s3, s3-init;
|
||
# queue (internal) — api, worker, broker. Наружу (127.0.0.1) опубликован только LB.
|
||
name: messenger-lab6
|
||
|
||
# lab5: общие настройки очереди для API (relay) и worker.
|
||
x-queue-env: &queue-env
|
||
BROKER_URL_FILE: /run/secrets/broker_url
|
||
BROKER_STREAM: ${BROKER_STREAM:-attachments.jobs}
|
||
BROKER_GROUP: ${BROKER_GROUP:-attachment-workers}
|
||
BROKER_TIMEOUT_SECONDS: ${BROKER_TIMEOUT_SECONDS:-2}
|
||
PIPELINE_VERSION: ${PIPELINE_VERSION:-1}
|
||
JOB_MAX_ATTEMPTS: ${JOB_MAX_ATTEMPTS:-5}
|
||
JOB_TIMEOUT_SECONDS: ${JOB_TIMEOUT_SECONDS:-60}
|
||
JOB_LEASE_GRACE_SECONDS: ${JOB_LEASE_GRACE_SECONDS:-10}
|
||
JOB_CLAIM_IDLE_SECONDS: ${JOB_CLAIM_IDLE_SECONDS:-75}
|
||
JOB_MAX_DELIVERIES: ${JOB_MAX_DELIVERIES:-10}
|
||
JOB_RETRY_BASE_SECONDS: ${JOB_RETRY_BASE_SECONDS:-2}
|
||
JOB_RETRY_MAX_SECONDS: ${JOB_RETRY_MAX_SECONDS:-60}
|
||
STAGING_RETENTION_SECONDS: ${STAGING_RETENTION_SECONDS:-86400}
|
||
RELAY_POLL_INTERVAL_SECONDS: ${RELAY_POLL_INTERVAL_SECONDS:-1}
|
||
RELAY_FAST_PATH_GRACE_SECONDS: ${RELAY_FAST_PATH_GRACE_SECONDS:-2}
|
||
|
||
# lab6: OpenTelemetry SDK (src/telemetry.py). Очереди SDK ограничены: при недоступном
|
||
# Collector батчи отбрасываются, RAM процесса не растёт.
|
||
x-otel-env: &otel-env
|
||
OTEL_SDK_DISABLED: ${OTEL_SDK_DISABLED:-false}
|
||
OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://otel-collector:4318}
|
||
OTEL_EXPORTER_OTLP_PROTOCOL: http/protobuf
|
||
# Python SDK: секунды
|
||
OTEL_EXPORTER_OTLP_TIMEOUT: "5"
|
||
OTEL_RESOURCE_ATTRIBUTES: ${OTEL_RESOURCE_ATTRIBUTES:-deployment.environment.name=lab}
|
||
OTEL_SEMCONV_STABILITY_OPT_IN: http
|
||
OTEL_BSP_MAX_QUEUE_SIZE: "2048"
|
||
OTEL_BSP_MAX_EXPORT_BATCH_SIZE: "512"
|
||
OTEL_BSP_EXPORT_TIMEOUT: "5000"
|
||
OTEL_BLRP_MAX_QUEUE_SIZE: "2048"
|
||
OTEL_BLRP_EXPORT_TIMEOUT: "5000"
|
||
OTEL_METRIC_EXPORT_INTERVAL: "10000"
|
||
|
||
x-api-env: &api-env
|
||
APP_ENV: ${APP_ENV:-lab}
|
||
APP_VERSION: ${APP_VERSION:-lab6}
|
||
OTEL_SERVICE_NAME: messenger-api
|
||
<<: [*queue-env, *otel-env]
|
||
SESSION_STORE: postgres
|
||
SESSION_TTL_SECONDS: ${SESSION_TTL_SECONDS:-3600}
|
||
SESSION_CACHE_TTL_SECONDS: ${SESSION_CACHE_TTL_SECONDS:-60}
|
||
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_POOL_TIMEOUT_SECONDS: ${DB_POOL_TIMEOUT_SECONDS:-3}
|
||
DB_CONNECT_TIMEOUT_SECONDS: ${DB_CONNECT_TIMEOUT_SECONDS:-3}
|
||
DB_STATEMENT_TIMEOUT_MS: ${DB_STATEMENT_TIMEOUT_MS:-5000}
|
||
CACHE_TIMEOUT_SECONDS: ${CACHE_TIMEOUT_SECONDS:-0.25}
|
||
CACHE_RETRY_AFTER_SECONDS: ${CACHE_RETRY_AFTER_SECONDS:-5}
|
||
IDEMPOTENCY_TTL_SECONDS: ${IDEMPOTENCY_TTL_SECONDS:-86400}
|
||
READY_DB_TIMEOUT_SECONDS: ${READY_DB_TIMEOUT_SECONDS:-1}
|
||
DRAIN_SECONDS: ${DRAIN_SECONDS:-3}
|
||
GRACEFUL_TIMEOUT_SECONDS: ${GRACEFUL_TIMEOUT_SECONDS:-10}
|
||
KEEP_ALIVE_SECONDS: ${KEEP_ALIVE_SECONDS:-30}
|
||
PASSWORD_HASH_CONCURRENCY: ${PASSWORD_HASH_CONCURRENCY:-4}
|
||
# Реплики НЕ применяют миграции — это делает сервис migrate.
|
||
RUN_MIGRATIONS_ON_START: "false"
|
||
# lab4: S3 (credentials API — отдельные, только Read/Write/List одного бакета)
|
||
S3_ENDPOINT: ${S3_ENDPOINT:-http://s3:8333}
|
||
S3_REGION: ${S3_REGION:-us-east-1}
|
||
S3_BUCKET: ${S3_BUCKET:-messenger-attachments}
|
||
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_access_key_id
|
||
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_secret_access_key
|
||
S3_CONNECT_TIMEOUT_SECONDS: ${S3_CONNECT_TIMEOUT_SECONDS:-2}
|
||
S3_READ_TIMEOUT_SECONDS: ${S3_READ_TIMEOUT_SECONDS:-10}
|
||
S3_MAX_ATTEMPTS: ${S3_MAX_ATTEMPTS:-2}
|
||
# Общий бюджет вызова S3 (вкл. DNS) и circuit breaker — см. src/main.py s3_call
|
||
S3_DOWNLOAD_DEADLINE_SECONDS: ${S3_DOWNLOAD_DEADLINE_SECONDS:-5}
|
||
S3_UPLOAD_DEADLINE_SECONDS: ${S3_UPLOAD_DEADLINE_SECONDS:-10}
|
||
S3_CIRCUIT_OPEN_SECONDS: ${S3_CIRCUIT_OPEN_SECONDS:-5}
|
||
# Лимиты загрузки (не выше контракта: 10 MiB, 40 Мп)
|
||
MAX_UPLOAD_BYTES: ${MAX_UPLOAD_BYTES:-10485760}
|
||
MAX_IMAGE_PIXELS: ${MAX_IMAGE_PIXELS:-40000000}
|
||
MAX_CONCURRENT_UPLOADS: ${MAX_CONCURRENT_UPLOADS:-4}
|
||
UPLOAD_QUEUE_TIMEOUT_SECONDS: ${UPLOAD_QUEUE_TIMEOUT_SECONDS:-5}
|
||
UPLOAD_READ_TIMEOUT_SECONDS: ${UPLOAD_READ_TIMEOUT_SECONDS:-60}
|
||
IMAGE_DECODE_CONCURRENCY: ${IMAGE_DECODE_CONCURRENCY:-2}
|
||
# Сборщик незавершённых загрузок и объектов-сирот (src/reconcile.py)
|
||
ATTACHMENT_GC_INTERVAL_SECONDS: ${ATTACHMENT_GC_INTERVAL_SECONDS:-300}
|
||
ATTACHMENT_GC_GRACE_SECONDS: ${ATTACHMENT_GC_GRACE_SECONDS:-900}
|
||
ORPHAN_SWEEP_INTERVAL_SECONDS: ${ORPHAN_SWEEP_INTERVAL_SECONDS:-3600}
|
||
|
||
x-api-healthcheck: &api-healthcheck
|
||
test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8080/health/ready', timeout=2).read()"]
|
||
interval: 3s
|
||
timeout: 2s
|
||
retries: 2
|
||
start_period: 20s
|
||
|
||
x-api-limits: &api-limits
|
||
limits:
|
||
cpus: "0.50"
|
||
# lab4: 4 одновременные загрузки по 10 MiB; замер пика — evidence/upload_memory*.json
|
||
memory: 320M
|
||
|
||
x-api-secrets: &api-secrets
|
||
- database_url
|
||
- cache_url
|
||
- s3_access_key_id
|
||
- s3_secret_access_key
|
||
- broker_url
|
||
|
||
services:
|
||
postgres:
|
||
image: postgres:16.4-alpine
|
||
restart: unless-stopped
|
||
environment:
|
||
POSTGRES_USER: messenger
|
||
POSTGRES_DB: messenger
|
||
POSTGRES_PASSWORD_FILE: /run/secrets/postgres_password
|
||
# (API_REPLICAS + 1 при rolling update) x DB_POOL_MAX + migrate + запас < 100
|
||
command: ["postgres", "-c", "max_connections=100"]
|
||
secrets:
|
||
- postgres_password
|
||
volumes:
|
||
- pgdata:/var/lib/postgresql/data
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "pg_isready -U messenger -d messenger -h 127.0.0.1"]
|
||
interval: 3s
|
||
timeout: 3s
|
||
retries: 20
|
||
start_period: 10s
|
||
networks: [backend]
|
||
stop_grace_period: 30s
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "1.00"
|
||
memory: 512M
|
||
|
||
redis:
|
||
image: redis:7.4.1-alpine
|
||
restart: unless-stopped
|
||
user: redis
|
||
# Только кэш проверок сессии: без RDB/AOF, ограничение памяти и LRU.
|
||
command:
|
||
- sh
|
||
- -c
|
||
- exec redis-server --requirepass "$$(tr -d '\r\n' < /run/secrets/redis_password)" --save "" --appendonly no --maxmemory 64mb --maxmemory-policy allkeys-lru
|
||
secrets:
|
||
- redis_password
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "REDISCLI_AUTH=\"$$(tr -d '\r\n' < /run/secrets/redis_password)\" redis-cli ping | grep -q PONG"]
|
||
interval: 3s
|
||
timeout: 3s
|
||
retries: 20
|
||
networks: [backend]
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.25"
|
||
memory: 96M
|
||
|
||
# lab5: брокер заданий — ОТДЕЛЬНЫЙ от кэша сессий Redis (тот же закреплённый образ).
|
||
# Кэш: без persistence и с вытеснением LRU; брокер: AOF с fsync на каждую запись
|
||
# (ответ на XADD = сообщение на диске, аналог publisher confirm) и noeviction
|
||
# (поток никогда не вытесняется). Volume brokerdata переживает рестарт/пересоздание.
|
||
broker:
|
||
image: redis:7.4.1-alpine
|
||
restart: unless-stopped
|
||
user: redis
|
||
command:
|
||
- sh
|
||
- -c
|
||
- exec redis-server --requirepass "$$(tr -d '\r\n' < /run/secrets/broker_password)" --dir /data --appendonly yes --appendfsync always --save "" --maxmemory 256mb --maxmemory-policy noeviction
|
||
secrets:
|
||
- broker_password
|
||
volumes:
|
||
- brokerdata:/data
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "REDISCLI_AUTH=\"$$(tr -d '\r\n' < /run/secrets/broker_password)\" redis-cli ping | grep -q PONG"]
|
||
interval: 3s
|
||
timeout: 3s
|
||
retries: 20
|
||
networks: [queue]
|
||
stop_grace_period: 10s
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 320M
|
||
|
||
# lab4: S3-совместимое хранилище SeaweedFS (master + volume + filer + S3 gateway в
|
||
# одном процессе `weed server`). Наружу НЕ публикуется. Административные HTTP-порты
|
||
# (master 9333, filer 8888, volume 8080) слушают только 127.0.0.1 ВНУТРИ контейнера
|
||
# (-ip.bind), в сеть storage выставлен только S3 API (8333). Анонимного доступа нет:
|
||
# identities из secrets/s3_config (bootstrap-admin и messenger-api с правами
|
||
# Read/Write/List только на свой бакет). Bucket создаёт s3-init.
|
||
s3:
|
||
image: chrislusf/seaweedfs:${SEAWEEDFS_VERSION:-4.50}
|
||
restart: unless-stopped
|
||
entrypoint: ["/usr/bin/weed"]
|
||
command:
|
||
- server
|
||
- -dir=/data
|
||
- -ip=127.0.0.1
|
||
- -ip.bind=127.0.0.1
|
||
- -s3
|
||
- -s3.port=8333
|
||
- -s3.ip.bind=0.0.0.0
|
||
- -s3.config=/run/secrets/s3_config
|
||
- -s3.iam=false
|
||
- -s3.port.iceberg=0
|
||
- -s3.port.lance=0
|
||
- -s3.autoCreateBucket=false
|
||
# lab5: без ограничения 8 одновременных PUT по 9.5 MiB (2 реплики x
|
||
# MAX_CONCURRENT_UPLOADS=4) доводили weed до лимита 512M -> OOM-kill и
|
||
# рестарт S3 (docker events: oom/die/start). Лишние загрузки ждут в S3.
|
||
- -s3.concurrentUploadLimitMB=128
|
||
- -master.volumeSizeLimitMB=256
|
||
- -master.electionTimeout=2s
|
||
- -master.telemetry=false
|
||
- -volume.max=0
|
||
# Go runtime не знает о лимите cgroup и при GOGC=100 копит мусор до ~2x живой
|
||
# кучи. Мягкий лимит заставляет GC работать раньше, чем придёт OOM killer.
|
||
environment:
|
||
GOMEMLIMIT: 400MiB
|
||
secrets:
|
||
- s3_config
|
||
volumes:
|
||
- s3data:/data
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "wget -q -O /dev/null http://127.0.0.1:8333/healthz"]
|
||
interval: 3s
|
||
timeout: 3s
|
||
retries: 20
|
||
start_period: 10s
|
||
networks: [storage]
|
||
# Штатная остановка SeaweedFS длится ~20 с; Docker ждёт 10 с и шлёт SIGKILL.
|
||
# Подтверждённые записи уже в page cache ОС, том консистентен (см. REPORT.md).
|
||
stop_grace_period: 10s
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "1.00"
|
||
memory: 512M
|
||
|
||
# Идемпотентный bootstrap бакета + проверка приватности и прав API-ключей.
|
||
# Повторный запуск: "already exists (idempotent no-op)". Admin-ключи есть только здесь.
|
||
s3-init:
|
||
build: .
|
||
command: ["python", "-m", "src.s3_bootstrap"]
|
||
restart: "no"
|
||
environment:
|
||
S3_ENDPOINT: ${S3_ENDPOINT:-http://s3:8333}
|
||
S3_REGION: ${S3_REGION:-us-east-1}
|
||
S3_BUCKET: ${S3_BUCKET:-messenger-attachments}
|
||
S3_ADMIN_ACCESS_KEY_ID_FILE: /run/secrets/s3_admin_access_key_id
|
||
S3_ADMIN_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_admin_secret_access_key
|
||
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_access_key_id
|
||
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_secret_access_key
|
||
secrets:
|
||
- s3_admin_access_key_id
|
||
- s3_admin_secret_access_key
|
||
- s3_access_key_id
|
||
- s3_secret_access_key
|
||
depends_on:
|
||
s3:
|
||
condition: service_healthy
|
||
networks: [storage]
|
||
healthcheck:
|
||
disable: true
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 192M
|
||
|
||
# Однократные миграции до старта реплик. Повтор — no-op (schema_migrations),
|
||
# параллельный запуск сериализуется pg_advisory_lock.
|
||
migrate:
|
||
build: .
|
||
command: ["python", "-m", "src.migrate"]
|
||
restart: "no"
|
||
environment:
|
||
DATABASE_URL_FILE: /run/secrets/database_url
|
||
DB_CONNECT_TIMEOUT_SECONDS: ${DB_CONNECT_TIMEOUT_SECONDS:-3}
|
||
# DDL может идти дольше обычного запроса API
|
||
DB_STATEMENT_TIMEOUT_MS: "60000"
|
||
MIGRATE_WAIT_SECONDS: "60"
|
||
secrets:
|
||
- database_url
|
||
depends_on:
|
||
postgres:
|
||
condition: service_healthy
|
||
networks: [backend]
|
||
healthcheck:
|
||
disable: true
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 256M
|
||
|
||
api:
|
||
build: .
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
restart: unless-stopped
|
||
environment: *api-env
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
postgres:
|
||
condition: service_healthy
|
||
redis:
|
||
condition: service_healthy
|
||
migrate:
|
||
condition: service_completed_successfully
|
||
s3-init:
|
||
condition: service_completed_successfully
|
||
# Порт только внутри сети backend — в обход LB API недоступен.
|
||
expose:
|
||
- "8080"
|
||
healthcheck: *api-healthcheck
|
||
# SIGTERM -> осушение DRAIN_SECONDS (3 с) -> graceful shutdown (<= 10 с).
|
||
stop_signal: SIGTERM
|
||
stop_grace_period: 20s
|
||
read_only: true
|
||
tmpfs:
|
||
- /tmp:size=16m
|
||
networks: [backend, storage, queue, telemetry]
|
||
# Остановленный s3 исчезает из Docker DNS; короткий таймаут резолвера.
|
||
dns_opt: ["timeout:1", "attempts:1"]
|
||
deploy:
|
||
replicas: ${API_REPLICAS:-2}
|
||
resources: *api-limits
|
||
|
||
# lab5: worker — отдельный процесс без HTTP. Берёт задания из broker, пишет
|
||
# final/ в S3 и результат в PostgreSQL. Также крутит outbox relay (sweeper +
|
||
# сборщик просроченных аренд) и сборщик staging/сирот: публикация не зависит от
|
||
# того, живы ли реплики API. Недоступность worker НЕ делает API синхронным.
|
||
worker:
|
||
# Тот же образ, что у API (один Dockerfile); build здесь, чтобы `up --build`
|
||
# не пытался скачать локальный тег из registry.
|
||
build: .
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
command: ["python", "-m", "src.worker"]
|
||
restart: unless-stopped
|
||
environment:
|
||
<<: [*queue-env, *otel-env]
|
||
OTEL_SERVICE_NAME: messenger-worker
|
||
APP_VERSION: ${APP_VERSION:-lab6}
|
||
APP_ENV: ${APP_ENV:-lab}
|
||
DATABASE_URL_FILE: /run/secrets/database_url
|
||
DB_POOL_MIN: "1"
|
||
DB_POOL_MAX: ${WORKER_DB_POOL_MAX:-6}
|
||
DB_POOL_TIMEOUT_SECONDS: ${DB_POOL_TIMEOUT_SECONDS:-3}
|
||
DB_CONNECT_TIMEOUT_SECONDS: ${DB_CONNECT_TIMEOUT_SECONDS:-3}
|
||
DB_STATEMENT_TIMEOUT_MS: ${DB_STATEMENT_TIMEOUT_MS:-5000}
|
||
S3_ENDPOINT: ${S3_ENDPOINT:-http://s3:8333}
|
||
S3_REGION: ${S3_REGION:-us-east-1}
|
||
S3_BUCKET: ${S3_BUCKET:-messenger-attachments}
|
||
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_access_key_id
|
||
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_secret_access_key
|
||
S3_CONNECT_TIMEOUT_SECONDS: ${S3_CONNECT_TIMEOUT_SECONDS:-2}
|
||
S3_READ_TIMEOUT_SECONDS: ${S3_READ_TIMEOUT_SECONDS:-10}
|
||
S3_MAX_ATTEMPTS: ${S3_MAX_ATTEMPTS:-2}
|
||
S3_CIRCUIT_OPEN_SECONDS: ${S3_CIRCUIT_OPEN_SECONDS:-5}
|
||
MAX_UPLOAD_BYTES: ${MAX_UPLOAD_BYTES:-10485760}
|
||
MAX_IMAGE_PIXELS: ${MAX_IMAGE_PIXELS:-40000000}
|
||
WORKER_CONCURRENCY: ${WORKER_CONCURRENCY:-2}
|
||
WORKER_IMAGE_TIMEOUT_SECONDS: ${WORKER_IMAGE_TIMEOUT_SECONDS:-30}
|
||
WORKER_IMAGE_MEMORY_MB: ${WORKER_IMAGE_MEMORY_MB:-1024}
|
||
ATTACHMENT_GC_INTERVAL_SECONDS: ${ATTACHMENT_GC_INTERVAL_SECONDS:-300}
|
||
ATTACHMENT_GC_GRACE_SECONDS: ${ATTACHMENT_GC_GRACE_SECONDS:-900}
|
||
ORPHAN_SWEEP_INTERVAL_SECONDS: ${ORPHAN_SWEEP_INTERVAL_SECONDS:-3600}
|
||
WORKER_FAULTS: "false"
|
||
secrets:
|
||
- database_url
|
||
- s3_access_key_id
|
||
- s3_secret_access_key
|
||
- broker_url
|
||
depends_on:
|
||
postgres:
|
||
condition: service_healthy
|
||
broker:
|
||
condition: service_healthy
|
||
migrate:
|
||
condition: service_completed_successfully
|
||
s3-init:
|
||
condition: service_completed_successfully
|
||
healthcheck:
|
||
# Heartbeat обновляется, только пока все consumer-потоки проходят цикл.
|
||
test: ["CMD", "python", "-c", "import os,time,sys; sys.exit(0 if time.time()-os.path.getmtime('/tmp/worker-heartbeat') < 90 else 1)"]
|
||
interval: 5s
|
||
timeout: 3s
|
||
retries: 3
|
||
start_period: 20s
|
||
# SIGTERM: не брать новые сообщения, дождаться текущих (< JOB_TIMEOUT_SECONDS).
|
||
stop_signal: SIGTERM
|
||
stop_grace_period: 75s
|
||
read_only: true
|
||
tmpfs:
|
||
# forkserver-сокет и heartbeat; изображения в памяти, не на диске
|
||
- /tmp:size=16m
|
||
networks: [backend, storage, queue, telemetry]
|
||
dns_opt: ["timeout:1", "attempts:1"]
|
||
deploy:
|
||
replicas: ${WORKER_REPLICAS:-1}
|
||
resources:
|
||
limits:
|
||
cpus: "1.00"
|
||
# 2 подпроцесса обработки (RLIMIT_AS 1024 MiB виртуальной памяти, реальный
|
||
# пик 361 MiB на 40 Мп RGBA PNG, замер в REPORT.md) + процесс worker
|
||
memory: 1280M
|
||
|
||
# Экземпляр с коротким TTL — только для автотеста истечения сессии (профиль ttl).
|
||
# Не входит в LB (другое DNS-имя), портов наружу не публикует: тест запускается
|
||
# внутри сети командой `make test-expiry`
|
||
api-short-ttl:
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
profiles: [ttl]
|
||
restart: "no"
|
||
environment:
|
||
<<: *api-env
|
||
SESSION_TTL_SECONDS: ${SHORT_SESSION_TTL_SECONDS:-2}
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
expose:
|
||
- "8080"
|
||
healthcheck: *api-healthcheck
|
||
stop_grace_period: 20s
|
||
read_only: true
|
||
tmpfs:
|
||
- /tmp:size=16m
|
||
networks: [backend, storage, queue, telemetry]
|
||
deploy:
|
||
resources: *api-limits
|
||
|
||
# lab4: экземпляры с ВНЕДРЁННЫМИ отказами пути загрузки (профиль faults), только для
|
||
# student-tests/test_attachment_consistency.py. Не входят в LB (другие DNS-имена),
|
||
# порты не публикуют. Запуск: python scripts/ops.py in-network consistency
|
||
api-fault-s3:
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
profiles: [faults]
|
||
restart: "no"
|
||
environment:
|
||
<<: *api-env
|
||
FAULT_INJECTION: s3_error_on_put
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
expose: ["8080"]
|
||
read_only: true
|
||
tmpfs: ["/tmp:size=16m"]
|
||
networks: [backend, storage, queue, telemetry]
|
||
deploy:
|
||
resources: *api-limits
|
||
|
||
api-fault-db:
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
profiles: [faults]
|
||
restart: "no"
|
||
environment:
|
||
<<: *api-env
|
||
FAULT_INJECTION: db_error_after_s3_put
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
expose: ["8080"]
|
||
read_only: true
|
||
tmpfs: ["/tmp:size=16m"]
|
||
networks: [backend, storage, queue, telemetry]
|
||
deploy:
|
||
resources: *api-limits
|
||
|
||
# lab5: разрыв COMMIT -> publish. Реплика отвечает 202 (задание и outbox уже
|
||
# в PostgreSQL), затем в background task до XADD завершает процесс (os._exit).
|
||
# Восстановление: sweeper любой другой реплики/worker публикует строку outbox.
|
||
api-fault-publish:
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
profiles: [faults]
|
||
restart: "no"
|
||
environment:
|
||
<<: *api-env
|
||
FAULT_INJECTION: crash_after_commit
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
expose: ["8080"]
|
||
read_only: true
|
||
tmpfs: ["/tmp:size=16m"]
|
||
networks: [backend, storage, queue, telemetry]
|
||
deploy:
|
||
resources: *api-limits
|
||
|
||
# lab5: разрыв publish -> подтверждение в outbox: XADD выполнен, процесс умирает до
|
||
# DELETE строки outbox -> повторная публикация (дубль в брокере) -> worker no-op.
|
||
api-fault-dup:
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
profiles: [faults]
|
||
restart: "no"
|
||
environment:
|
||
<<: *api-env
|
||
FAULT_INJECTION: crash_after_publish
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
expose: ["8080"]
|
||
read_only: true
|
||
tmpfs: ["/tmp:size=16m"]
|
||
networks: [backend, storage, queue, telemetry]
|
||
deploy:
|
||
resources: *api-limits
|
||
|
||
# Процесс завершается (os._exit) между S3 PUT (staging) и COMMIT — «падение» реплики.
|
||
api-fault-crash:
|
||
image: messenger-api:${APP_VERSION:-lab6}
|
||
profiles: [faults]
|
||
restart: "no"
|
||
environment:
|
||
<<: *api-env
|
||
FAULT_INJECTION: crash_after_s3_put
|
||
secrets: *api-secrets
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
expose: ["8080"]
|
||
read_only: true
|
||
tmpfs: ["/tmp:size=16m"]
|
||
networks: [backend, storage, queue, telemetry]
|
||
deploy:
|
||
resources: *api-limits
|
||
|
||
lb:
|
||
image: haproxy:2.8.30-alpine
|
||
restart: unless-stopped
|
||
ports:
|
||
- "127.0.0.1:${APP_PORT:-8080}:8080"
|
||
- "127.0.0.1:${LB_STATS_PORT:-8404}:8404"
|
||
volumes:
|
||
- ./infra/haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro
|
||
depends_on:
|
||
api:
|
||
condition: service_healthy
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "wget -q -O /dev/null http://127.0.0.1:8404/lb-health"]
|
||
interval: 5s
|
||
timeout: 2s
|
||
retries: 3
|
||
stop_grace_period: 5s
|
||
networks: [backend]
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 128M
|
||
|
||
# ------------------------------------------------------------------ lab6: наблюдаемость
|
||
otel-collector:
|
||
image: otel/opentelemetry-collector-contrib:0.115.1
|
||
restart: unless-stopped
|
||
command: ["--config=/etc/otelcol/collector.yaml"]
|
||
volumes:
|
||
- ./infra/otel/collector.yaml:/etc/otelcol/collector.yaml:ro
|
||
environment:
|
||
# Go runtime: мягкий лимит ниже лимита контейнера (memory_limiter — 180 MiB)
|
||
GOMEMLIMIT: 200MiB
|
||
networks: [telemetry]
|
||
stop_grace_period: 10s
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
# Замер: пик RSS 234 MiB при недоступных Tempo+Loki и ~300 RPS (очереди 20 батчей полны).
|
||
memory: 384M
|
||
|
||
prometheus:
|
||
image: prom/prometheus:v2.55.1
|
||
restart: unless-stopped
|
||
command:
|
||
- --config.file=/etc/prometheus/prometheus.yml
|
||
- --storage.tsdb.path=/prometheus
|
||
- --storage.tsdb.retention.time=7d
|
||
- --web.enable-lifecycle
|
||
volumes:
|
||
- ./infra/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
|
||
- ./infra/prometheus/rules:/etc/prometheus/rules:ro
|
||
- promdata:/prometheus
|
||
ports:
|
||
# Только чтение состояния alerts/запросов с хоста: http://127.0.0.1:9090/alerts
|
||
- "127.0.0.1:${PROMETHEUS_PORT:-9090}:9090"
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "wget -q -O /dev/null http://127.0.0.1:9090/-/ready"]
|
||
interval: 5s
|
||
timeout: 3s
|
||
retries: 20
|
||
networks: [telemetry]
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 512M
|
||
|
||
loki:
|
||
image: grafana/loki:3.3.2
|
||
restart: unless-stopped
|
||
command: ["-config.file=/etc/loki/loki.yaml"]
|
||
volumes:
|
||
- ./infra/loki/loki.yaml:/etc/loki/loki.yaml:ro
|
||
- lokidata:/loki
|
||
networks: [telemetry]
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 512M
|
||
|
||
# Образ Tempo работает от uid 10001, а новый named volume принадлежит root:
|
||
# однократно отдаём каталог пользователю tempo (как в примерах Grafana).
|
||
tempo-init:
|
||
image: grafana/tempo:2.6.1
|
||
user: root
|
||
entrypoint: ["chown", "10001:10001", "/var/tempo"]
|
||
restart: "no"
|
||
volumes:
|
||
- tempodata:/var/tempo
|
||
network_mode: none
|
||
|
||
tempo:
|
||
image: grafana/tempo:2.6.1
|
||
restart: unless-stopped
|
||
command: ["-config.file=/etc/tempo/tempo.yaml"]
|
||
volumes:
|
||
- ./infra/tempo/tempo.yaml:/etc/tempo/tempo.yaml:ro
|
||
- tempodata:/var/tempo
|
||
depends_on:
|
||
tempo-init:
|
||
condition: service_completed_successfully
|
||
networks: [telemetry]
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 512M
|
||
|
||
grafana:
|
||
image: grafana/grafana:11.4.0
|
||
restart: unless-stopped
|
||
environment:
|
||
GF_SECURITY_ADMIN_USER: admin
|
||
GF_SECURITY_ADMIN_PASSWORD__FILE: /run/secrets/grafana_admin_password
|
||
GF_USERS_ALLOW_SIGN_UP: "false"
|
||
GF_AUTH_ANONYMOUS_ENABLED: "false"
|
||
GF_ANALYTICS_REPORTING_ENABLED: "false"
|
||
GF_ANALYTICS_CHECK_FOR_UPDATES: "false"
|
||
GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH: /var/lib/grafana/dashboards/messenger.json
|
||
secrets:
|
||
- grafana_admin_password
|
||
volumes:
|
||
- ./infra/grafana/provisioning:/etc/grafana/provisioning:ro
|
||
- ./infra/grafana/dashboards:/var/lib/grafana/dashboards:ro
|
||
- grafanadata:/var/lib/grafana
|
||
ports:
|
||
- "127.0.0.1:${GRAFANA_PORT:-3000}:3000"
|
||
depends_on:
|
||
prometheus:
|
||
condition: service_healthy
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "wget -q -O /dev/null http://127.0.0.1:3000/api/health"]
|
||
interval: 5s
|
||
timeout: 3s
|
||
retries: 30
|
||
networks: [telemetry]
|
||
deploy:
|
||
resources:
|
||
limits:
|
||
cpus: "0.50"
|
||
memory: 256M
|
||
|
||
volumes:
|
||
pgdata:
|
||
s3data:
|
||
brokerdata:
|
||
# lab6: история телеметрии переживает пересоздание контейнеров
|
||
promdata:
|
||
lokidata:
|
||
tempodata:
|
||
grafanadata:
|
||
|
||
networks:
|
||
backend:
|
||
# Внутренняя сеть хранилища объектов: без выхода наружу и без LB.
|
||
storage:
|
||
internal: true
|
||
# lab5: сеть брокера — только API (relay), worker и broker.
|
||
queue:
|
||
internal: true
|
||
# lab6: OTLP и хранилища телеметрии. Не internal: Grafana/Prometheus публикуются
|
||
# на 127.0.0.1; OTLP-порты Collector и порты Loki/Tempo наружу не публикуются.
|
||
telemetry:
|
||
|
||
secrets:
|
||
postgres_password:
|
||
file: ./secrets/postgres_password
|
||
redis_password:
|
||
file: ./secrets/redis_password
|
||
database_url:
|
||
file: ./secrets/database_url
|
||
cache_url:
|
||
file: ./secrets/cache_url
|
||
s3_config:
|
||
file: ./secrets/s3_config.json
|
||
s3_admin_access_key_id:
|
||
file: ./secrets/s3_admin_access_key_id
|
||
s3_admin_secret_access_key:
|
||
file: ./secrets/s3_admin_secret_access_key
|
||
s3_access_key_id:
|
||
file: ./secrets/s3_access_key_id
|
||
s3_secret_access_key:
|
||
file: ./secrets/s3_secret_access_key
|
||
broker_password:
|
||
file: ./secrets/broker_password
|
||
broker_url:
|
||
file: ./secrets/broker_url
|
||
grafana_admin_password:
|
||
file: ./secrets/grafana_admin_password
|