docker-compose.yml

services:
  postgres:
    image: postgres:16-alpine
    container_name: stellar_alerts_postgres
    restart: unless-stopped
    environment:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: postgrespassword
      POSTGRES_DB: stellar_alerts
    ports:
      - "5432:5432"
    volumes:
      - postgres_data:/var/lib/postgresql/data
    healthcheck:
      test: [ "CMD-SHELL", "pg_isready -U postgres -d stellar_alerts" ]
      interval: 5s
      timeout: 5s
      retries: 5

  redis:
    image: redis:7-alpine
    container_name: stellar_alerts_redis
    restart: unless-stopped
    ports:
      - "6379:6379"
    healthcheck:
      test: [ "CMD-SHELL", "redis-cli ping | grep PONG" ]
      interval: 5s
      timeout: 5s
      retries: 5

  # Chaos-engineering fault injection proxy (#145). The API test suite spins
  # up its own local fixture server on an ephemeral port and talks to it
  # *through* toxiproxy so tests can inject deterministic latency/timeouts/
  # disconnects — see apps/api/src/__tests__/chaos.test.ts. Not used at
  # runtime by the app itself, only by the chaos test suite (gated on
  # TOXIPROXY_URL).
  #
  # Uses host networking so toxiproxy can reach whatever ephemeral port the
  # test process's fixture server bound to on localhost, without needing to
  # pre-declare/pass that port into the container. This is Linux-only
  # (matches the ubuntu-latest CI runner this is primarily meant for); on
  # macOS/Windows Docker Desktop, run toxiproxy directly on the host instead
  # (`docker run --network host ghcr.io/shopify/toxiproxy:latest` won't work
  # there either — use `-p 8474:8474 -p 8666:8666` and point
  # TOXIPROXY_UPSTREAM_HOST at host.docker.internal).
  toxiproxy:
    image: ghcr.io/shopify/toxiproxy:latest
    container_name: stellar_alerts_toxiproxy
    restart: unless-stopped
    ports:
      - "8474:8474"
      - "8666:8666"
    extra_hosts:
      - "host.docker.internal:host-gateway"

volumes:
  postgres_data: