Files
mtusi2026lab5/compose.yaml
T
zemiles 13bec4ab8b ЛР 5 — БПИ2303 — Лазарев — на 4
Асинхронная обработка вложений: RabbitMQ 4.1 (quorum-очереди, publisher
confirms, manual ack, DLQ), transactional outbox, отдельный worker.

- upload: оригинал в staging S3, затем одной транзакцией queued + jobs +
  outbox -> 202; API к брокеру не подключается
- relay: публикация outbox с confirm, reaper истёкших аренд, redrive
- worker: claim по строке jobs, аренда с heartbeat и fencing по
  lease_token, transform в подпроцессе с RLIMIT_AS/RLIMIT_CPU/таймаутом,
  thumbnail 256, crop 128, WebP 1280 по детерминированным ключам
- retry с backoff и jitter, постоянные/временные ошибки, реестр failed,
  ops replay (ready не повторяется), DLQ для невалидных сообщений
- janitor: очистка staging и сирот; миграция 004 с backfill вложений lab4
- тесты отказов: окна COMMIT/publish, крахи worker до/после S3 и ack,
  заморозка worker, всплеск загрузок, права брокера; evidence/ и REPORT.md
2026-11-13 16:28:18 +03:00

398 lines
15 KiB
YAML
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Lab5: HAProxy -> API x N (по умолчанию 2) -> PostgreSQL (данные, сессии, вложения, jobs, outbox)
# -> SeaweedFS S3 (staging/ и attachments/, приватный bucket)
# relay: outbox (PostgreSQL) -> RabbitMQ (quorum queue, publisher confirms) + reaper аренд
# worker x N: RabbitMQ -> staging -> transform (thumbnail/crop/optimized) -> S3 + PostgreSQL
# Запуск: ./scripts/init-secrets.sh && docker compose --env-file .env up -d --build --wait
name: mtusi-lab5
# Один образ на восемь сервисов. Собирает его ТОЛЬКО migrate (первый job стенда), остальные
# используют готовый тег: параллельная сборка одного тега несколькими сервисами давала
# «image ... already exists» на пустом Docker.
x-app-image: &app-image
image: mtusi-messenger:lab5
pull_policy: never
x-db-env: &db-env
POSTGRES_HOST: db
POSTGRES_PORT: "5432"
POSTGRES_DB: ${POSTGRES_DB:-messenger}
POSTGRES_USER: ${POSTGRES_USER:-messenger}
POSTGRES_PASSWORD_FILE: /run/secrets/postgres_password
x-s3-env: &s3-env
S3_ENDPOINT: ${S3_ENDPOINT:-http://s3:8333}
S3_REGION: ${S3_REGION:-us-east-1}
S3_BUCKET: ${S3_BUCKET:-messenger-attachments}
S3_CONNECT_TIMEOUT_SECONDS: ${S3_CONNECT_TIMEOUT_SECONDS:-2}
S3_READ_TIMEOUT_SECONDS: ${S3_READ_TIMEOUT_SECONDS:-5}
x-job-env: &job-env
PIPELINE_VERSION: ${PIPELINE_VERSION:-1}
JOB_MAX_ATTEMPTS: ${JOB_MAX_ATTEMPTS:-5}
JOB_TIMEOUT_SECONDS: ${JOB_TIMEOUT_SECONDS:-60}
JOB_LEASE_SECONDS: ${JOB_LEASE_SECONDS:-30}
JOB_RETRY_BASE_SECONDS: ${JOB_RETRY_BASE_SECONDS:-2}
JOB_RETRY_MAX_SECONDS: ${JOB_RETRY_MAX_SECONDS:-60}
MAX_UPLOAD_BYTES: ${MAX_UPLOAD_BYTES:-10485760}
MAX_IMAGE_PIXELS: ${MAX_IMAGE_PIXELS:-40000000}
services:
db:
image: postgres:16.4-alpine@sha256:5660c2cbfea50c7a9127d17dc4e48543eedd3d7a41a595a2dfa572471e37e64c
environment:
POSTGRES_DB: ${POSTGRES_DB:-messenger}
POSTGRES_USER: ${POSTGRES_USER:-messenger}
POSTGRES_PASSWORD_FILE: /run/secrets/postgres_password
secrets: [postgres_password]
volumes:
- pgdata:/var/lib/postgresql/data
networks: [backend]
restart: unless-stopped
healthcheck:
test: ["CMD-SHELL", "pg_isready -U \"$${POSTGRES_USER}\" -d \"$${POSTGRES_DB}\""]
interval: 2s
timeout: 3s
retries: 30
stop_grace_period: 30s
deploy:
resources:
limits: {cpus: "1.0", memory: 512M}
# S3 (как в lab4): master/volume/filer только на 127.0.0.1 контейнера, наружу — S3 API :8333.
s3:
image: chrislusf/seaweedfs:4.50@sha256:c2658820e763294a9e9de7b36c81e59b879e3459148351aee4e31a126ca64860
command:
- server
- -dir=/data
- -ip=127.0.0.1
- -ip.bind=127.0.0.1
- -master.volumeSizeLimitMB=1024
- -master.electionTimeout=2s
- -volume.preStopSeconds=2
- -volume.max=0
- -s3
- -s3.ip.bind=0.0.0.0
- -s3.port=8333
- -s3.config=/run/secrets/s3_identities.json
- -s3.iam=false
- -s3.port.iceberg=0
- -s3.port.lance=0
- -s3.concurrentUploadLimitMB=64
- -filer.concurrentUploadLimitMB=64
- -volume.concurrentUploadLimitMB=64
- -s3.cacheCapacityMB=16
- -s3.readerCacheSizeMB=32
secrets: [s3_identities.json]
environment:
GOMEMLIMIT: 512MiB
volumes:
- s3data:/data
networks: [backend]
restart: unless-stopped
healthcheck:
test: ["CMD-SHELL", "wget -q -O /dev/null http://127.0.0.1:8333/healthz && curl -sf -m 2 -X PUT --data-binary ok http://127.0.0.1:8888/.healthcheck/probe -o /dev/null"]
interval: 5s
start_interval: 1s
timeout: 3s
retries: 3
start_period: 60s
stop_grace_period: 30s
deploy:
resources:
limits: {cpus: "1.0", memory: 768M}
# Брокер: RabbitMQ 4.1, quorum-очередь (durable, Raft-журнал на диске), publisher confirms.
# Топология и пользователи — из definitions (secret). Порты на хост не публикуются.
broker:
image: rabbitmq:4.1.4-alpine@sha256:b736d649308e1b3e1a116c3f36986b605ee3d03e88f10166be2900083d2e63f2
hostname: broker # имя узла rabbit@broker неизменно -> данные volume находятся после пересоздания
volumes:
- brokerdata:/var/lib/rabbitmq
- ./infra/rabbitmq/rabbitmq.conf:/etc/rabbitmq/conf.d/90-lab5.conf:ro
secrets:
- source: rabbitmq_definitions.json
target: /run/secrets/rabbitmq_definitions.json
networks: [backend]
restart: unless-stopped
healthcheck:
# Узел запущен, нет алармов памяти/диска, и рабочая очередь объявлена.
test: ["CMD-SHELL", "rabbitmq-diagnostics -q check_running && rabbitmq-diagnostics -q check_local_alarms && rabbitmqctl -q -p messenger list_queues name | grep -q attachments.process.v1"]
interval: 10s
start_interval: 2s
timeout: 15s
retries: 5
start_period: 60s
stop_grace_period: 30s
deploy:
resources:
limits: {cpus: "1.0", memory: 512M}
migrate:
<<: *app-image
build:
context: .
args:
APP_VERSION: ${APP_VERSION:-1}
pull_policy: build
command: ["python", "/app/src/migrate.py"]
environment:
<<: *db-env
secrets: [postgres_password]
networks: [backend]
healthcheck: {disable: true}
restart: "no"
depends_on:
db: {condition: service_healthy}
deploy:
resources:
limits: {cpus: "0.5", memory: 128M}
s3-init:
<<: *app-image
command: ["python", "/app/src/s3_bootstrap.py"]
environment:
<<: *s3-env
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_admin_access_key
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_admin_secret_key
secrets: [s3_admin_access_key, s3_admin_secret_key]
networks: [backend]
healthcheck: {disable: true}
restart: "no"
depends_on:
s3: {condition: service_healthy}
deploy:
resources:
limits: {cpus: "0.5", memory: 128M}
api:
<<: *app-image
expose: ["8080"]
environment:
<<: [*db-env, *s3-env, *job-env]
APP_HOST: 0.0.0.0
APP_PORT: "8080"
APP_ENV: ${APP_ENV:-lab}
SESSION_TTL_SECONDS: ${SESSION_TTL_SECONDS:-3600}
IDEMPOTENCY_TTL_SECONDS: ${IDEMPOTENCY_TTL_SECONDS:-86400}
DB_POOL_SIZE: ${DB_POOL_SIZE:-10}
DRAIN_SECONDS: ${DRAIN_SECONDS:-3}
SHUTDOWN_TIMEOUT_SECONDS: ${SHUTDOWN_TIMEOUT_SECONDS:-10}
PASSWORD_HASH_CONCURRENCY: ${PASSWORD_HASH_CONCURRENCY:-2}
PASSWORD_HASH_WAIT_SECONDS: ${PASSWORD_HASH_WAIT_SECONDS:-0.5}
DB_HEALTH_INTERVAL_SECONDS: ${DB_HEALTH_INTERVAL_SECONDS:-1}
READY_STALE_SECONDS: ${READY_STALE_SECONDS:-3}
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_api_access_key
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_api_secret_key
UPLOAD_TIMEOUT_SECONDS: ${UPLOAD_TIMEOUT_SECONDS:-60}
UPLOAD_CONCURRENCY: ${UPLOAD_CONCURRENCY:-8}
DECODE_CONCURRENCY: ${DECODE_CONCURRENCY:-2}
S3_PUT_CONCURRENCY: ${S3_PUT_CONCURRENCY:-3}
FAULT_INJECTION: ${FAULT_INJECTION:-0}
# У API НЕТ доступа к брокеру: задание пишется в outbox PostgreSQL.
secrets: [postgres_password, s3_api_access_key, s3_api_secret_key]
networks: [backend]
restart: unless-stopped
stop_grace_period: 20s
depends_on:
db: {condition: service_healthy}
migrate: {condition: service_completed_successfully}
s3-init: {condition: service_completed_successfully}
deploy:
replicas: ${API_REPLICAS:-2}
resources:
limits: {cpus: "0.5", memory: 256M}
# Outbox relay + reaper аренд (src/relay.py). Можно масштабировать: SKIP LOCKED + advisory lock.
relay:
<<: *app-image
command: ["python", "/app/src/relay.py"]
environment:
<<: [*db-env, *job-env]
SERVICE_NAME: relay
BROKER_URL_FILE: /run/secrets/broker_relay_url
RELAY_POLL_SECONDS: ${RELAY_POLL_SECONDS:-1}
REAPER_INTERVAL_SECONDS: ${REAPER_INTERVAL_SECONDS:-5}
JOB_REDRIVE_AFTER_SECONDS: ${JOB_REDRIVE_AFTER_SECONDS:-900}
OUTBOX_RETENTION_SECONDS: ${OUTBOX_RETENTION_SECONDS:-86400}
RELAY_FAULT: ${RELAY_FAULT:-}
secrets: [postgres_password, broker_relay_url]
networks: [backend]
healthcheck:
test: ["CMD", "python", "-c", "import os,sys,time; sys.exit(0 if time.time() - os.path.getmtime('/tmp/relay.heartbeat') < 30 else 1)"]
interval: 10s
timeout: 3s
retries: 3
start_period: 30s
restart: unless-stopped
stop_grace_period: 15s
depends_on:
migrate: {condition: service_completed_successfully}
broker: {condition: service_healthy}
deploy:
resources:
limits: {cpus: "0.25", memory: 128M}
# Worker (src/worker.py): WORKER_CONCURRENCY потоков x prefetch 1 на контейнер.
worker:
<<: *app-image
command: ["python", "/app/src/worker.py"]
environment:
<<: [*db-env, *s3-env, *job-env]
SERVICE_NAME: worker
BROKER_URL_FILE: /run/secrets/broker_worker_url
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_worker_access_key
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_worker_secret_key
# Чтение 10 MiB из staging и запись результатов: дольше, чем upload-таймауты API.
S3_READ_TIMEOUT_SECONDS: ${WORKER_S3_READ_TIMEOUT_SECONDS:-10}
WORKER_CONCURRENCY: ${WORKER_CONCURRENCY:-2}
WORKER_PREFETCH: ${WORKER_PREFETCH:-1}
TRANSFORM_MEMORY_LIMIT_MB: ${TRANSFORM_MEMORY_LIMIT_MB:-768}
TRANSFORM_CPU_SECONDS: ${TRANSFORM_CPU_SECONDS:-30}
WORKER_TMP_DIR: /tmp/work
WORKER_FAULT_INJECTION: ${WORKER_FAULT_INJECTION:-0}
FAULT_SLOW_SECONDS: ${FAULT_SLOW_SECONDS:-20}
secrets: [postgres_password, broker_worker_url, s3_worker_access_key, s3_worker_secret_key]
networks: [backend]
healthcheck:
test: ["CMD", "python", "-c", "import os,sys,time; sys.exit(0 if time.time() - os.path.getmtime('/tmp/worker.heartbeat') < 30 else 1)"]
interval: 10s
timeout: 3s
retries: 3
start_period: 30s
restart: unless-stopped
# SIGTERM: потребители дорабатывают текущее задание (<= JOB_TIMEOUT 60 с) и выходят.
stop_grace_period: 75s
depends_on:
migrate: {condition: service_completed_successfully}
s3-init: {condition: service_completed_successfully}
broker: {condition: service_healthy}
deploy:
replicas: ${WORKER_REPLICAS:-2}
resources:
limits: {cpus: "1.0", memory: 1G}
janitor:
<<: *app-image
command: ["python", "/app/src/janitor.py"]
environment:
<<: [*db-env, *s3-env]
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_janitor_access_key
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_janitor_secret_key
ATTACHMENT_GRACE_SECONDS: ${ATTACHMENT_GRACE_SECONDS:-3600}
JANITOR_INTERVAL_SECONDS: ${JANITOR_INTERVAL_SECONDS:-300}
STAGING_RETENTION_SECONDS: ${STAGING_RETENTION_SECONDS:-86400}
secrets: [postgres_password, s3_janitor_access_key, s3_janitor_secret_key]
networks: [backend]
healthcheck:
test: ["CMD", "python", "-c", "import os,sys,time; sys.exit(0 if time.time() - os.path.getmtime('/tmp/janitor.heartbeat') < 90 else 1)"]
interval: 15s
timeout: 3s
retries: 2
start_period: 20s
restart: unless-stopped
depends_on:
migrate: {condition: service_completed_successfully}
s3-init: {condition: service_completed_successfully}
deploy:
resources:
limits: {cpus: "0.25", memory: 128M}
lb:
image: haproxy:3.0.5-alpine@sha256:43a2591a6a62a41dd22edb4452880e9c70de489fbd7c54ffb52d035cccbfe949
volumes:
- ./infra/haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro
environment:
LOGIN_RATE_LIMIT: ${LOGIN_RATE_LIMIT:-20}
ports:
- "127.0.0.1:${LB_PORT:-8080}:8080"
- "127.0.0.1:8404:8404"
networks: [frontend, backend]
restart: unless-stopped
healthcheck:
test: ["CMD", "haproxy", "-c", "-q", "-f", "/usr/local/etc/haproxy/haproxy.cfg"]
interval: 10s
timeout: 3s
retries: 3
depends_on:
api: {condition: service_started}
deploy:
resources:
limits: {cpus: "0.5", memory: 128M}
# Операторский CLI (не часть стенда): docker compose --profile tools run --rm ops <команда>
# inspect / failed / replay / dlq. Ключ S3 janitor (List, без чтения), брокер — пользователь ops.
ops:
<<: *app-image
profiles: [tools]
entrypoint: ["python", "/app/src/ops.py"]
environment:
<<: [*db-env, *s3-env, *job-env]
SERVICE_NAME: ops
BROKER_URL_FILE: /run/secrets/broker_ops_url
S3_ACCESS_KEY_ID_FILE: /run/secrets/s3_janitor_access_key
S3_SECRET_ACCESS_KEY_FILE: /run/secrets/s3_janitor_secret_key
secrets: [postgres_password, broker_ops_url, s3_janitor_access_key, s3_janitor_secret_key]
networks: [backend]
healthcheck: {disable: true}
restart: "no"
# Проверки внутри backend-сети (Pillow для попиксельной проверки вариантов, права S3).
checks:
<<: *app-image
profiles: [tools]
entrypoint: ["python"]
environment:
<<: *s3-env
volumes:
- ./student-tests:/app/student-tests:ro
- ./tests:/app/tests:ro
- ./contracts:/app/contracts:ro
- ./lab.json:/app/lab.json:ro
- ./evidence:/app/evidence
secrets: [s3_api_access_key, s3_api_secret_key, s3_worker_access_key, s3_worker_secret_key,
s3_janitor_access_key, s3_janitor_secret_key]
networks: [backend]
healthcheck: {disable: true}
restart: "no"
user: "0"
secrets:
postgres_password:
file: ./secrets/postgres_password
s3_identities.json:
file: ./secrets/s3_identities.json
s3_admin_access_key:
file: ./secrets/s3_admin_access_key
s3_admin_secret_key:
file: ./secrets/s3_admin_secret_key
s3_api_access_key:
file: ./secrets/s3_api_access_key
s3_api_secret_key:
file: ./secrets/s3_api_secret_key
s3_worker_access_key:
file: ./secrets/s3_worker_access_key
s3_worker_secret_key:
file: ./secrets/s3_worker_secret_key
s3_janitor_access_key:
file: ./secrets/s3_janitor_access_key
s3_janitor_secret_key:
file: ./secrets/s3_janitor_secret_key
rabbitmq_definitions.json:
file: ./secrets/rabbitmq_definitions.json
broker_relay_url:
file: ./secrets/broker_relay_url
broker_worker_url:
file: ./secrets/broker_worker_url
broker_ops_url:
file: ./secrets/broker_ops_url
networks:
frontend: {}
backend:
internal: true
volumes:
pgdata: {}
s3data: {}
brokerdata: {}