forked from mtusi2026/mtusi2026lab5
- upload возвращает 202 после записи оригинала в staging/ и одной транзакции attachments + jobs + outbox - relay outbox → RabbitMQ (quorum-очередь, publisher confirms) в каждой реплике API; sweeper возвращает работу с истёкшей арендой - worker: аренда с fencing-токеном, варианты thumbnail/crop/optimized (sharp), retry с backoff и jitter, реестр failed, replay с аудитом, reprocess новой версией pipeline - reconciler дочищает staging/ и variants/ - сценарии уровней 3–4, fault-injection трека 2 (14 сценариев, инварианты I1–I6), evidence и REPORT.md Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
258 lines
11 KiB
YAML
258 lines
11 KiB
YAML
# Клиент → lb (HAProxy) → api × N → postgres (метаданные, jobs, outbox) + s3 (SeaweedFS, staging/ оригинала).
|
||
# api (relay outbox) → rabbitmq (quorum-очередь attachments.process) → worker × M → s3 (originals/, variants/) + postgres.
|
||
# Запуск: scripts/init-secrets.sh && docker compose --env-file .env up -d --build (см. REPORT.md)
|
||
name: messenger-lab5
|
||
|
||
x-app: &app
|
||
build: .
|
||
image: messenger-lab5-app:dev # уникальное имя: на общем Docker-host не перепутать с чужими стендами
|
||
networks: [backend]
|
||
|
||
x-s3-env: &s3-env
|
||
S3_ENDPOINT: ${S3_ENDPOINT:-http://s3:9000}
|
||
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
|
||
|
||
x-jobs-env: &jobs-env
|
||
DATABASE_URL_FILE: /run/secrets/database_url
|
||
BROKER_URL_FILE: /run/secrets/broker_url
|
||
MAX_IMAGE_PIXELS: ${MAX_IMAGE_PIXELS:-40000000}
|
||
PIPELINE_VERSION: ${PIPELINE_VERSION:-1}
|
||
JOB_MAX_ATTEMPTS: ${JOB_MAX_ATTEMPTS:-5}
|
||
JOB_TIMEOUT_SECONDS: ${JOB_TIMEOUT_SECONDS:-60}
|
||
JOB_RETRY_BASE_MS: ${JOB_RETRY_BASE_MS:-2000}
|
||
JOB_REPUBLISH_SECONDS: ${JOB_REPUBLISH_SECONDS:-600}
|
||
FAILPOINTS: ${FAILPOINTS:-0} # 1 только на тестовом стенде: файл failpoint-<точка>.png один раз роняет процесс
|
||
|
||
x-seaweedfs: &seaweedfs
|
||
image: chrislusf/seaweedfs:4.48@sha256:4e61d15fd35994cb1e43e1e553dff106794841fd9a99ade2fc8c8bfce4d7872d
|
||
restart: unless-stopped
|
||
stop_grace_period: 20s
|
||
|
||
services:
|
||
postgres:
|
||
image: postgres:18.6-alpine
|
||
command: ["postgres", "-c", "max_connections=200", "-c", "shared_buffers=256MB"]
|
||
environment:
|
||
POSTGRES_USER: messenger
|
||
POSTGRES_DB: messenger
|
||
POSTGRES_PASSWORD_FILE: /run/secrets/postgres_password
|
||
secrets: [postgres_password]
|
||
volumes: [pgdata:/var/lib/postgresql]
|
||
healthcheck:
|
||
test: ["CMD-SHELL", "pg_isready -U messenger -d messenger"]
|
||
interval: 2s
|
||
timeout: 2s
|
||
retries: 15
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "${DB_CPUS:-2}", memory: 1g }
|
||
stop_grace_period: 30s
|
||
restart: unless-stopped
|
||
networks: [backend]
|
||
|
||
# Хранилище SeaweedFS: master + volume + filer, данные в volume s3data. Только во внутренней сети storage:
|
||
# его HTTP-интерфейсы (master :9333, filer :8888, volume :8080) не требуют подписи S3 и недоступны даже из API.
|
||
seaweedfs:
|
||
<<: *seaweedfs
|
||
command: ["server", "-dir=/data", "-ip=seaweedfs", "-ip.bind=0.0.0.0", "-filer", "-master.volumeSizeLimitMB=1024"]
|
||
volumes: [s3data:/data]
|
||
healthcheck:
|
||
test: ["CMD", "wget", "-qO", "/dev/null", "http://127.0.0.1:8888/"]
|
||
interval: 2s
|
||
timeout: 2s
|
||
retries: 30
|
||
start_period: 5s
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "1", memory: 512m }
|
||
networks: [storage]
|
||
|
||
# S3-шлюз — единственный путь к байтам: проверяет SigV4 и права identities (secrets/s3_identities.json).
|
||
# Встроенный IAM API, Iceberg и Lance выключены; порт не опубликован на хост.
|
||
s3:
|
||
<<: *seaweedfs
|
||
# concurrentUploadLimitMB: шлюз буферизует чанки (4 MiB) каждой загрузки; сверх лимита новые загрузки ждут,
|
||
# а не выводят шлюз за лимит памяти (эксперимент research/upload_memory.py: при 256m без лимита — OOM).
|
||
command: ["s3", "-filer=seaweedfs:8888", "-port=9000", "-config=/run/secrets/s3_identities",
|
||
"-iam=false", "-port.iceberg=0", "-port.lance=0", "-concurrentUploadLimitMB=256"]
|
||
secrets: [s3_identities]
|
||
depends_on:
|
||
seaweedfs: { condition: service_healthy }
|
||
healthcheck:
|
||
test: ["CMD", "wget", "-qO", "/dev/null", "http://127.0.0.1:9000/healthz"]
|
||
interval: 2s
|
||
timeout: 2s
|
||
retries: 30
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "1", memory: 768m }
|
||
networks: [storage, backend]
|
||
|
||
# Одноразовые сервисы до старта реплик: схема БД (advisory lock) и bucket (повторный запуск ничего не меняет).
|
||
migrate:
|
||
<<: *app
|
||
command: ["node", "src/main.ts", "migrate"]
|
||
environment:
|
||
DATABASE_URL_FILE: /run/secrets/database_url
|
||
secrets: [database_url]
|
||
restart: "no"
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "0.5", memory: 128m }
|
||
depends_on:
|
||
postgres: { condition: service_healthy }
|
||
|
||
s3-init:
|
||
<<: *app
|
||
command: ["node", "src/main.ts", "s3-init"]
|
||
environment: *s3-env
|
||
secrets:
|
||
- { source: s3_admin_access_key_id, target: s3_access_key_id }
|
||
- { source: s3_admin_secret_access_key, target: s3_secret_access_key }
|
||
restart: "no"
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "0.5", memory: 128m }
|
||
depends_on:
|
||
s3: { condition: service_healthy }
|
||
|
||
# Брокер: RabbitMQ, quorum-очередь на диске (volume rabbitmq-data), publisher confirms, ручной ack.
|
||
# hostname фиксирует имя узла rabbit@rabbitmq: данные в volume привязаны к нему и переживают пересоздание.
|
||
rabbitmq:
|
||
image: rabbitmq:4.3.6-alpine@sha256:6931816af1a943f471c038b278835d36692d55dec769232739296bb91b7de0f5
|
||
hostname: rabbitmq
|
||
volumes:
|
||
- rabbitmq-data:/var/lib/rabbitmq
|
||
- ./infra/rabbitmq/rabbitmq.conf:/etc/rabbitmq/conf.d/90-messenger.conf:ro
|
||
secrets: [rabbitmq_definitions]
|
||
healthcheck:
|
||
test: ["CMD", "rabbitmq-diagnostics", "-q", "check_port_connectivity"]
|
||
interval: 5s
|
||
timeout: 10s
|
||
retries: 30
|
||
start_period: 20s
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "1", memory: 512m }
|
||
stop_grace_period: 30s
|
||
restart: unless-stopped
|
||
networks: [backend]
|
||
|
||
api:
|
||
<<: *app
|
||
environment:
|
||
<<: [*s3-env, *jobs-env]
|
||
SESSION_TTL_SECONDS: ${SESSION_TTL_SECONDS:-3600}
|
||
DB_POOL_MAX: ${DB_POOL_MAX:-8}
|
||
SHUTDOWN_DRAIN_MS: ${SHUTDOWN_DRAIN_MS:-2500}
|
||
MAX_UPLOAD_BYTES: ${MAX_UPLOAD_BYTES:-10485760}
|
||
secrets:
|
||
- database_url
|
||
- { source: broker_url_api, target: broker_url }
|
||
- { source: s3_api_access_key_id, target: s3_access_key_id }
|
||
- { source: s3_api_secret_access_key, target: s3_secret_access_key }
|
||
# rabbitmq не в depends_on: API принимает загрузки и без брокера, задания ждут в outbox.
|
||
depends_on:
|
||
migrate: { condition: service_completed_successfully }
|
||
s3-init: { condition: service_completed_successfully }
|
||
healthcheck:
|
||
test: ["CMD", "wget", "-qO", "/dev/null", "http://127.0.0.1:8080/health/ready"]
|
||
interval: 2s
|
||
timeout: 2s
|
||
retries: 3
|
||
start_period: 10s
|
||
deploy:
|
||
replicas: ${API_REPLICAS:-2}
|
||
resources:
|
||
limits: { cpus: "${API_CPUS:-1}", memory: "${API_MEMORY:-256m}" }
|
||
stop_grace_period: 15s
|
||
restart: unless-stopped # упавшая реплика поднимается сама; после `docker stop` остаётся остановленной
|
||
|
||
# Worker: отдельный процесс без HTTP, конкурирующие consumers одной очереди. WORKER_CONCURRENCY = prefetch =
|
||
# одновременных заданий на реплику; память ограничена контейнером, libvips — один поток, без кэша.
|
||
worker:
|
||
<<: *app
|
||
command: ["node", "src/main.ts", "worker"]
|
||
environment:
|
||
<<: [*s3-env, *jobs-env]
|
||
WORKER_CONCURRENCY: ${WORKER_CONCURRENCY:-2}
|
||
secrets:
|
||
- database_url
|
||
- { source: broker_url_worker, target: broker_url }
|
||
- { source: s3_worker_access_key_id, target: s3_access_key_id }
|
||
- { source: s3_worker_secret_access_key, target: s3_secret_access_key }
|
||
depends_on:
|
||
migrate: { condition: service_completed_successfully }
|
||
s3-init: { condition: service_completed_successfully }
|
||
rabbitmq: { condition: service_healthy }
|
||
deploy:
|
||
replicas: ${WORKER_REPLICAS:-2}
|
||
resources:
|
||
limits: { cpus: "${WORKER_CPUS:-1}", memory: "${WORKER_MEMORY:-768m}" }
|
||
stop_grace_period: 15s # SIGTERM: consumer отменяется, текущие задания до 10 с, затем выход
|
||
restart: unless-stopped
|
||
|
||
# Периодическая сверка БД ↔ S3 с grace period (src/reconcile.ts). Свои S3-ключи: объекты + листинг bucket.
|
||
reconciler:
|
||
<<: *app
|
||
command: ["node", "src/main.ts", "reconcile", "--apply", "--every=${RECONCILE_INTERVAL_SECONDS:-300}"]
|
||
environment:
|
||
<<: *s3-env
|
||
DATABASE_URL_FILE: /run/secrets/database_url
|
||
RECONCILE_GRACE_SECONDS: ${RECONCILE_GRACE_SECONDS:-900}
|
||
STAGING_RETENTION_SECONDS: ${STAGING_RETENTION_SECONDS:-86400}
|
||
secrets:
|
||
- database_url
|
||
- { source: s3_reconciler_access_key_id, target: s3_access_key_id }
|
||
- { source: s3_reconciler_secret_access_key, target: s3_secret_access_key }
|
||
depends_on:
|
||
migrate: { condition: service_completed_successfully }
|
||
s3-init: { condition: service_completed_successfully }
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "0.5", memory: 128m }
|
||
restart: unless-stopped
|
||
|
||
lb:
|
||
image: haproxy:3.2.26-alpine
|
||
volumes: ["./infra/haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro"]
|
||
ports:
|
||
- "127.0.0.1:${LB_PORT:-8080}:8080"
|
||
- "127.0.0.1:${LB_STATS_PORT:-8404}:8404" # статистика HAProxy для оператора, это не API
|
||
depends_on:
|
||
api: { condition: service_healthy }
|
||
deploy:
|
||
resources:
|
||
limits: { cpus: "${LB_CPUS:-1}", memory: 128m }
|
||
restart: unless-stopped
|
||
networks: [frontend, backend]
|
||
|
||
networks:
|
||
frontend: {}
|
||
backend: { internal: true } # api, worker, postgres, rabbitmq и s3-шлюз недоступны с хоста и не имеют выхода наружу
|
||
storage: { internal: true } # внутренние компоненты SeaweedFS видит только s3-шлюз
|
||
|
||
volumes:
|
||
pgdata: {}
|
||
s3data: {}
|
||
rabbitmq-data: {}
|
||
|
||
secrets:
|
||
database_url: { file: ./secrets/database_url }
|
||
postgres_password: { file: ./secrets/postgres_password }
|
||
s3_identities: { file: ./secrets/s3_identities.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_api_access_key_id: { file: ./secrets/s3_api_access_key_id }
|
||
s3_api_secret_access_key: { file: ./secrets/s3_api_secret_access_key }
|
||
s3_reconciler_access_key_id: { file: ./secrets/s3_reconciler_access_key_id }
|
||
s3_reconciler_secret_access_key: { file: ./secrets/s3_reconciler_secret_access_key }
|
||
s3_worker_access_key_id: { file: ./secrets/s3_worker_access_key_id }
|
||
s3_worker_secret_access_key: { file: ./secrets/s3_worker_secret_access_key }
|
||
rabbitmq_definitions: { file: ./secrets/rabbitmq_definitions.json }
|
||
broker_url_api: { file: ./secrets/broker_url_api }
|
||
broker_url_worker: { file: ./secrets/broker_url_worker }
|