chatballs/compose.yaml

# Канонический production-манифест Chatballs (ADR-HUB-0028).
# Один Compose для: установки клиента, staging, smoke.
# Локальная разработка — через override compose.dev.yaml; этот файл не содержит
# dev source mounts, dev-команд и host-портов (кроме public 80/443 у gateway).
#
# Image references берутся из release.env (immutable digest refs, см. ADR §release).
# В dev они не задаются — срабатывают безопасные defaults, а override подменяет
# image на build.
#
# Запуск production/box:
#   docker compose --env-file .env --env-file release.env up -d
# Запуск разработки:
#   docker compose -f compose.yaml -f compose.dev.yaml --env-file .env.example --env-file .env up

services:
  postgres:
    image: ${CHATBALLS_POSTGRES_IMAGE:-pgvector/pgvector:pg16}
    restart: unless-stopped
    environment:
      POSTGRES_DB: ${POSTGRES_DB:?POSTGRES_DB is required}
      POSTGRES_USER: ${POSTGRES_USER:?POSTGRES_USER is required}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:?POSTGRES_PASSWORD is required}
      POSTGRES_APP_USER: ${POSTGRES_APP_USER:?POSTGRES_APP_USER is required}
      POSTGRES_APP_PASSWORD: ${POSTGRES_APP_PASSWORD:?POSTGRES_APP_PASSWORD is required}
      POSTGRES_PLATFORM_USER: ${POSTGRES_PLATFORM_USER:?POSTGRES_PLATFORM_USER is required}
      POSTGRES_PLATFORM_PASSWORD: ${POSTGRES_PLATFORM_PASSWORD:?POSTGRES_PLATFORM_PASSWORD is required}
      POSTGRES_MIGRATION_USER: ${POSTGRES_MIGRATION_USER:?POSTGRES_MIGRATION_USER is required}
      POSTGRES_MIGRATION_PASSWORD: ${POSTGRES_MIGRATION_PASSWORD:?POSTGRES_MIGRATION_PASSWORD is required}
    volumes:
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/postgres:/var/lib/postgresql/data
      - ${CHATBALLS_RELEASE_DIR:-.}/deploy/postgres/init-runtime-roles.sh:/docker-entrypoint-initdb.d/20-chatballs-runtime-roles.sh:ro
      - ${CHATBALLS_RELEASE_DIR:-.}/deploy/postgres/reassign-schema-ownership.sql:/chatballs-reassign-ownership.sql:ro
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U $${POSTGRES_USER} -d $${POSTGRES_DB}"]
      interval: 10s
      timeout: 5s
      retries: 10

  redis:
    image: ${CHATBALLS_REDIS_IMAGE:-redis:7-alpine}
    restart: unless-stopped
    command: ["redis-server", "--appendonly", "yes"]
    volumes:
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/redis:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 20s

  # One-shot init: только миграции (ADR-HUB-0028 §one-shot init). Владельца и
  # демо-данные создаёт мастер первого запуска в браузере (/api/v1/setup/).
  # collectstatic испечён в backend-образ, здесь не вызывается.
  init:
    image: ${CHATBALLS_BACKEND_IMAGE:-chatballs-backend:dev}
    env_file:
      - ${CHATBALLS_INSTANCE_DIR:-.}/.env
    command: ["python", "manage.py", "migrate", "--noinput"]
    environment:
      CHATBALLS_DB_ROLE: migration
    restart: "no"
    depends_on:
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy

  backend-app:
    image: ${CHATBALLS_BACKEND_IMAGE:-chatballs-backend:dev}
    restart: unless-stopped
    env_file:
      - ${CHATBALLS_INSTANCE_DIR:-.}/.env
    command: >
      sh -c "gunicorn chatballs_backend.asgi_app:application
      --worker-class uvicorn.workers.UvicornWorker
      --bind 0.0.0.0:8000
      --workers $${CHATBALLS_GUNICORN_WORKERS:-3}
      --timeout $${CHATBALLS_GUNICORN_TIMEOUT:-60}"
    environment:
      CHATBALLS_DB_ROLE: app
    volumes:
      # Legacy source media retained for the separately approved copy/hash
      # migration. Production writes use the required S3 backend in C04.
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/media:/app/apps/backend/media
    depends_on:
      init:
        condition: service_completed_successfully
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
    healthcheck:
      test:
        [
          "CMD",
          "python",
          "-c",
          "import os, urllib.request; req = urllib.request.Request('http://127.0.0.1:8000/api/v1/health/live/', headers={'Host': os.environ.get('CHATBALLS_APP_HEALTHCHECK_HOST', 'app.localhost'), 'X-Forwarded-Proto': 'https'}); urllib.request.urlopen(req, timeout=3)",
        ]
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 20s

  backend-platform:
    image: ${CHATBALLS_BACKEND_IMAGE:-chatballs-backend:dev}
    restart: unless-stopped
    env_file:
      - ${CHATBALLS_INSTANCE_DIR:-.}/.env
    command: >
      sh -c "gunicorn chatballs_backend.asgi_platform:application
      --worker-class uvicorn.workers.UvicornWorker
      --bind 0.0.0.0:8000
      --workers $${CHATBALLS_GUNICORN_WORKERS:-3}
      --timeout $${CHATBALLS_GUNICORN_TIMEOUT:-60}"
    environment:
      CHATBALLS_DB_ROLE: platform
    depends_on:
      init:
        condition: service_completed_successfully
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
    healthcheck:
      test:
        [
          "CMD",
          "python",
          "-c",
          "import os, urllib.request; req = urllib.request.Request('http://127.0.0.1:8000/api/v1/health/live/', headers={'Host': os.environ.get('CHATBALLS_PLATFORM_HEALTHCHECK_HOST', 'platform.localhost'), 'X-Forwarded-Proto': 'https'}); urllib.request.urlopen(req, timeout=3)",
        ]
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 20s

  # Django admin не подключён к gateway. Единственный host bind — loopback;
  # удалённый доступ допускается только через SSH tunnel (ADR-HUB-0031).
  backend-admin:
    image: ${CHATBALLS_BACKEND_IMAGE:-chatballs-backend:dev}
    restart: unless-stopped
    env_file:
      - ${CHATBALLS_INSTANCE_DIR:-.}/.env
    command: >
      sh -c "gunicorn chatballs_backend.asgi_admin:application
      --worker-class uvicorn.workers.UvicornWorker
      --bind 0.0.0.0:8000
      --workers $${CHATBALLS_GUNICORN_WORKERS:-2}
      --timeout $${CHATBALLS_GUNICORN_TIMEOUT:-60}"
    # Break-glass technical surface only: schema credentials are never used by
    # public app/platform runtimes and this service remains loopback-only.
    environment:
      CHATBALLS_DB_ROLE: migration
    ports:
      - "127.0.0.1:${CHATBALLS_ADMIN_PORT:-18001}:8000"
    depends_on:
      init:
        condition: service_completed_successfully
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy

  worker:
    image: ${CHATBALLS_BACKEND_IMAGE:-chatballs-backend:dev}
    restart: unless-stopped
    env_file:
      - ${CHATBALLS_INSTANCE_DIR:-.}/.env
    command: ["python", "manage.py", "run_worker"]
    environment:
      CHATBALLS_DB_ROLE: app
    volumes:
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/media:/app/apps/backend/media
    depends_on:
      backend-app:
        condition: service_healthy

  # Frontend: один nginx-образ со static-сборкой internal-ui и web-chat,
  # внутренняя маршрутизация (/api/, /ws/, /chat-widget.js,
  # /chat/, /calls/) — в deploy/nginx/frontend.production.conf. Не публикует
  # host-порт: public boundary — gateway (Caddy).
  frontend:
    image: ${CHATBALLS_FRONTEND_IMAGE:-chatballs-frontend:dev}
    restart: unless-stopped
    depends_on:
      backend-app:
        condition: service_healthy

  # Gateway: единственный HTTP/HTTPS public boundary (ADR-HUB-0028 §gateway).
  # Caddy: TLS termination, ACME, HTTP->HTTPS redirect, WebSocket upgrade (native).
  # Маршрутизация публичных путей делегируется frontend (internal). Coturn не
  # проксируется через Caddy — отдельная network boundary (profile calls).
  gateway:
    image: ${CHATBALLS_GATEWAY_IMAGE:-caddy:2.8.4}
    restart: unless-stopped
    environment:
      CHATBALLS_APP_DOMAIN: ${CHATBALLS_APP_DOMAIN:?CHATBALLS_APP_DOMAIN is required}
      CHATBALLS_PLATFORM_DOMAIN: ${CHATBALLS_PLATFORM_DOMAIN:?CHATBALLS_PLATFORM_DOMAIN is required}
      CHATBALLS_ACME_EMAIL: ${CHATBALLS_ACME_EMAIL:?CHATBALLS_ACME_EMAIL is required}
    ports:
      - "${CHATBALLS_WEB_LISTENING_IP:-0.0.0.0}:80:80"
      - "${CHATBALLS_WEB_LISTENING_IP:-0.0.0.0}:443:443"
    volumes:
      - ${CHATBALLS_RELEASE_DIR:-.}/Caddyfile:/etc/caddy/Caddyfile:ro
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/caddy:/data
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/caddy-config:/config
    depends_on:
      frontend:
        condition: service_started
      backend-app:
        condition: service_healthy
      backend-platform:
        condition: service_healthy

  # Coturn — медиа-relay для TURN fallback (SPEC-HUB-0013 §11).
  # Опциональный profile `calls` (ADR-HUB-0028): включается через COMPOSE_PROFILES=calls.
  # Host networking: relay использует широкий UDP-диапазон и реальный внешний IP,
  # публикуется напрямую и не проходит через HTTP reverse proxy. credentials
  # выдаёт backend по общему CHATBALLS_CALL_TURN_SECRET (static-auth-secret).
  coturn:
    profiles: ["calls"]
    image: ${CHATBALLS_COTURN_IMAGE:-coturn/coturn:4.6}
    restart: unless-stopped
    network_mode: host
    command:
      - -n
      - --log-file=stdout
      - --no-cli
      - --fingerprint
      - --use-auth-secret
      - --static-auth-secret=${CHATBALLS_CALL_TURN_SECRET}
      - --realm=${CHATBALLS_CALL_TURN_REALM}
      # Весь TURN живёт на выделенном публичном IP (отдельный порт/NIC), Caddy — на
      # основном IP. Это освобождает 443 под TURN-over-TLS без конфликта с web.
      - --listening-ip=${CHATBALLS_TURN_LISTENING_IP}
      - --relay-ip=${CHATBALLS_TURN_EXTERNAL_IP}
      - --external-ip=${CHATBALLS_TURN_EXTERNAL_IP}
      - --listening-port=${CHATBALLS_TURN_LISTENING_PORT:-3478}
      # TURN-over-TLS на 443: проходит через VPN/строгие сети, где UDP и 3478 режут.
      - --tls-listening-port=${CHATBALLS_TURN_TLS_PORT:-443}
      - --cert=/etc/coturn/certs/fullchain.pem
      - --pkey=/etc/coturn/certs/privkey.pem
      - --min-port=${CHATBALLS_TURN_MIN_PORT:-49160}
      # Relay-диапазон UDP (1 порт на media-endpoint; ~2 порта на звонок).
      # 49160-49999 = 840 портов ≈ 420 одновременных TURN-звонков на хосте.
      # Расширяется через CHATBALLS_TURN_MIN_PORT/CHATBALLS_TURN_MAX_PORT; все порты диапазона
      # должны быть открыты на firewall и не пересекаться с ephemeral-диапазоном ОС.
      - --max-port=${CHATBALLS_TURN_MAX_PORT:-49999}
      # Запрет анонимного и внутрисетевого relay (SPEC §11: без пересечения с хостом).
      - --no-multicast-peers
      - --no-tcp-relay
      - --denied-peer-ip=10.0.0.0-10.255.255.255
      - --denied-peer-ip=172.16.0.0-172.31.255.255
      - --denied-peer-ip=192.168.0.0-192.168.255.255
      - --no-tlsv1
      - --no-tlsv1_1
    volumes:
      # LE-сертификат TURN-хоста (turns:), скопированный под uid coturn (nobody)
      # renewal deploy-hook'ом. См. docs по развёртыванию TURN-over-TLS.
      - ${CHATBALLS_INSTANCE_DIR:-.}/data/coturn-certs:/etc/coturn/certs:ro
    healthcheck:
      # Allocation smoke: STUN binding к собственному listener на выделенном IP.
      test: ["CMD", "turnutils_stunclient", "-p", "${CHATBALLS_TURN_LISTENING_PORT:-3478}", "${CHATBALLS_TURN_LISTENING_IP}"]
      interval: 30s
      timeout: 5s
      retries: 5
      start_period: 15s