docker-compose.failover-controller.yml

x-rust-common: &rust-common
  build:
    context: .
    dockerfile: control-plane/Dockerfile.dev.shared
  volumes:
    - ./control-plane/shared:/app/control-plane/shared
    - ./Cargo.toml:/app/Cargo.toml
    - ./Cargo.lock:/app/Cargo.lock
    - cargo_cache:/usr/local/cargo/registry
    - cargo_git:/usr/local/cargo/git
  networks:
    - csf-network
  restart: unless-stopped

services:
  postgres:
    image: postgres:16-alpine
    container_name: csf-postgres-failover
    environment:
      POSTGRES_USER: csf_user
      POSTGRES_PASSWORD: csf_password
      POSTGRES_DB: csf_core
    ports:
      - "5432:5432"
    volumes:
      - postgres_data:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U csf_user -d csf_core"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - csf-network

  etcd:
    image: gcr.io/etcd-development/etcd:v3.5.21
    container_name: csf-etcd-failover
    command:
      - etcd
      - --advertise-client-urls=http://etcd:2379
      - --listen-client-urls=http://0.0.0.0:2379
    ports:
      - "2379:2379"
    networks:
      - csf-network

  api-gateway:
    <<: *rust-common
    container_name: csf-api-gateway-failover
    environment:
      DATABASE_URL: postgres://csf_user:csf_password@postgres:5432/csf_core
      RUST_LOG: debug
      JWT_SECRET: ${JWT_SECRET:-dev_jwt_secret_change_in_production}
      RSA_KEY_SIZE: "2048"
      REGISTRY_SERVICE_URL: http://registry:8001
      SCHEDULER_SERVICE_URL: http://scheduler:8002
      VOLUME_MANAGER_URL: http://volume-manager:8003
      FAILOVER_CONTROLLER_URL: http://failover-controller:8004
    ports:
      - "8000:8000"
    depends_on:
      postgres:
        condition: service_healthy
    volumes:
      - ./control-plane/api-gateway:/app/control-plane/api-gateway
      - ./control-plane/shared:/app/control-plane/shared
      - ./Cargo.toml:/app/Cargo.toml
      - ./Cargo.lock:/app/Cargo.lock
      - cargo_cache:/usr/local/cargo/registry
      - cargo_git:/usr/local/cargo/git
      - cargo_target_api_gateway:/app/target
    command: cargo watch -x "run -p api-gateway"

  registry:
    <<: *rust-common
    container_name: csf-registry-failover
    environment:
      DATABASE_URL: postgres://csf_user:csf_password@postgres:5432/csf_core
      ETCD_ENDPOINTS: http://etcd:2379
      REGISTRY_PORT: "8001"
      RUST_LOG: debug
      SCHEDULER_SERVICE_URL: http://scheduler:8002
    ports:
      - "8001:8001"
    depends_on:
      postgres:
        condition: service_healthy
      etcd:
        condition: service_started
    volumes:
      - ./control-plane/registry:/app/control-plane/registry
      - ./control-plane/shared:/app/control-plane/shared
      - ./Cargo.toml:/app/Cargo.toml
      - ./Cargo.lock:/app/Cargo.lock
      - cargo_cache:/usr/local/cargo/registry
      - cargo_git:/usr/local/cargo/git
      - cargo_target_registry:/app/target
    command: cargo watch -x "run -p registry"

  scheduler:
    <<: *rust-common
    container_name: csf-scheduler-failover
    environment:
      DATABASE_URL: postgres://csf_user:csf_password@postgres:5432/csf_core
      ETCD_ENDPOINTS: http://etcd:2379
      SCHEDULER_PORT: "8002"
      RUST_LOG: debug
    ports:
      - "8002:8002"
    depends_on:
      postgres:
        condition: service_healthy
      etcd:
        condition: service_started
    volumes:
      - ./control-plane/scheduler:/app/control-plane/scheduler
      - ./control-plane/shared:/app/control-plane/shared
      - ./Cargo.toml:/app/Cargo.toml
      - ./Cargo.lock:/app/Cargo.lock
      - cargo_cache:/usr/local/cargo/registry
      - cargo_git:/usr/local/cargo/git
      - cargo_target_scheduler:/app/target
    command: cargo watch -x "run -p scheduler"

  volume-manager:
    <<: *rust-common
    container_name: csf-volume-manager-failover
    environment:
      DATABASE_URL: postgres://csf_user:csf_password@postgres:5432/csf_core
      ETCD_ENDPOINTS: http://etcd:2379
      VOLUME_MANAGER_PORT: "8003"
      RUST_LOG: debug
    ports:
      - "8003:8003"
    depends_on:
      postgres:
        condition: service_healthy
      etcd:
        condition: service_started
    volumes:
      - ./control-plane/volume-manager:/app/control-plane/volume-manager
      - ./control-plane/shared:/app/control-plane/shared
      - ./Cargo.toml:/app/Cargo.toml
      - ./Cargo.lock:/app/Cargo.lock
      - cargo_cache:/usr/local/cargo/registry
      - cargo_git:/usr/local/cargo/git
      - cargo_target_volume_manager:/app/target
    command: cargo watch -x "run -p volume-manager"

  failover-controller:
    <<: *rust-common
    container_name: csf-failover-controller
    environment:
      DATABASE_URL: postgres://csf_user:csf_password@postgres:5432/csf_core
      FAILOVER_CONTROLLER_PORT: "8004"
      SCHEDULER_SERVICE_URL: http://scheduler:8002
      VOLUME_MANAGER_URL: http://volume-manager:8003
      RUST_LOG: debug
    ports:
      - "8004:8004"
    depends_on:
      postgres:
        condition: service_healthy
      scheduler:
        condition: service_started
      volume-manager:
        condition: service_started
    volumes:
      - ./control-plane/failover-controller:/app/control-plane/failover-controller
      - ./control-plane/shared:/app/control-plane/shared
      - ./Cargo.toml:/app/Cargo.toml
      - ./Cargo.lock:/app/Cargo.lock
      - cargo_cache:/usr/local/cargo/registry
      - cargo_git:/usr/local/cargo/git
      - cargo_target_failover_controller:/app/target
    command: cargo watch -x "run -p failover-controller"
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8004/health"]
      interval: 30s
      timeout: 10s
      retries: 3

volumes:
  postgres_data:
    driver: local
  cargo_cache:
    driver: local
  cargo_git:
    driver: local
  cargo_target_api_gateway:
    driver: local
  cargo_target_registry:
    driver: local
  cargo_target_scheduler:
    driver: local
  cargo_target_volume_manager:
    driver: local
  cargo_target_failover_controller:
    driver: local

networks:
  csf-network:
    driver: bridge