solstice-ci/deploy/podman/compose.yml

# Podman Compose production stack for Solstice CI with Traefik + Let's Encrypt
# Usage:
#  1) cp .env.sample .env && edit values (ENV=staging|prod, secrets, email)
#  2) Ensure DNS A/AAAA for required hostnames pointing to this host:
#       - traefik.svc.${DOMAIN}
#       - api.${ENV}.${DOMAIN}
#       - grpc.${ENV}.${DOMAIN}
#       - forge.${ENV}.${DOMAIN}
#       - github.${ENV}.${DOMAIN}
#       - minio.svc.${DOMAIN}
#       - s3.svc.${DOMAIN}
#       - mq.svc.${DOMAIN}
#       - db.svc.${DOMAIN} (optional)
#  3) podman compose -f compose.yml up -d --build

name: solstice-ci

networks:
  core: {}

volumes:
  traefik-acme:
  postgres-data:
  rabbitmq-data:
  minio-data:
  runner-state:


services:
  traefik:
    image: docker.io/library/traefik:v3.6
    container_name: traefik
    restart: unless-stopped
    environment:
      DOCKER_API_VERSION: ${DOCKER_API_VERSION:-1.44}
    command:
      - --log.level=DEBUG
      - --accesslog=true
      - --api.dashboard=true
      - --providers.docker=true
      - --providers.docker.exposedbydefault=false
      - --entrypoints.web.address=:80
      - --entrypoints.web.http.redirections.entrypoint.to=websecure
      - --entrypoints.web.http.redirections.entrypoint.scheme=https
      - --entrypoints.websecure.address=:443
      - --entrypoints.websecure.transport.respondingTimeouts.readTimeout=0s
      - --entrypoints.websecure.transport.respondingTimeouts.writeTimeout=0s
      - --entrypoints.websecure.transport.respondingTimeouts.idleTimeout=360s
      - --certificatesresolvers.le.acme.email=${TRAEFIK_ACME_EMAIL}
      - --certificatesresolvers.le.acme.storage=/acme/acme.json
      - --certificatesresolvers.le.acme.httpchallenge=true
      - --certificatesresolvers.le.acme.httpchallenge.entrypoint=web
      - --serversTransport.insecureSkipVerify=true
      # Optional: override ACME CA server via .env (e.g., staging URL)
      - --certificatesresolvers.le.acme.caserver=${TRAEFIK_ACME_CASERVER:-}
    ports:
      # Rootless Podman cannot bind privileged ports (<1024). Use high ports via .env (e.g., 8080/4443),
      # or adjust sysctl on the host: net.ipv4.ip_unprivileged_port_start=80 (requires root).
      - ${TRAEFIK_HTTP_PORT:-80}:80
      - ${TRAEFIK_HTTPS_PORT:-443}:443
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:Z
      - traefik-acme:/acme
    networks:
      - core
    labels:
      - traefik.enable=true
      # Dashboard at traefik.svc.${DOMAIN}
      - traefik.http.routers.traefik.rule=Host(`traefik.svc.${DOMAIN}`)
      - traefik.http.routers.traefik.entrypoints=websecure
      - traefik.http.routers.traefik.tls.certresolver=le
      - traefik.http.routers.traefik.service=api@internal
      - traefik.http.middlewares.traefik-auth.basicauth.users=${TRAEFIK_DASHBOARD_AUTH}
      - traefik.http.routers.traefik.middlewares=traefik-auth

  postgres:
    image: docker.io/library/postgres:16-alpine
    container_name: solstice-postgres
    restart: unless-stopped
    environment:
      POSTGRES_USER: ${POSTGRES_USER}
      POSTGRES_DB: ${POSTGRES_DB}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER}"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    ports:
      - "127.0.0.1:5432:5432"   # expose Postgres to host only
    volumes:
      - postgres-data:/var/lib/postgresql/data:Z
    networks:
      - core

  postgres-setup:
    image: docker.io/library/postgres:16-alpine
    container_name: solstice-postgres-setup
    depends_on:
      postgres:
        condition: service_healthy
    entrypoint: ["/bin/sh", "-c"]
    environment:
      POSTGRES_USER: ${POSTGRES_USER}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
    command: >-
      "export PGPASSWORD=${POSTGRES_PASSWORD} &&
      psql -h postgres -U ${POSTGRES_USER} -v ON_ERROR_STOP=1 -tc \"SELECT 'CREATE DATABASE solstice_staging' WHERE NOT EXISTS (SELECT FROM pg_database WHERE datname='solstice_staging')\" | psql -h postgres -U ${POSTGRES_USER} &&
      psql -h postgres -U ${POSTGRES_USER} -v ON_ERROR_STOP=1 -tc \"SELECT 'CREATE DATABASE solstice_prod' WHERE NOT EXISTS (SELECT FROM pg_database WHERE datname='solstice_prod')\" | psql -h postgres -U ${POSTGRES_USER}"
    networks:
      - core

  rabbitmq:
    image: docker.io/library/rabbitmq:4-management-alpine
    container_name: solstice-rabbitmq
    restart: unless-stopped
    environment:
      RABBITMQ_DEFAULT_USER: ${RABBITMQ_DEFAULT_USER}
      RABBITMQ_DEFAULT_PASS: ${RABBITMQ_DEFAULT_PASS}
      RABBITMQ_DEFAULT_VHOST: solstice-${ENV}
    healthcheck:
      test: ["CMD", "rabbitmq-diagnostics", "-q", "ping"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 5s
    ports:
      - "127.0.0.1:5672:5672"   # expose AMQP to host only
    volumes:
      - rabbitmq-data:/var/lib/rabbitmq:Z
    networks:
      - core
    labels:
      - traefik.enable=true
      # Management UI at mq.svc.${DOMAIN}
      - traefik.http.routers.mq.rule=Host(`mq.svc.${DOMAIN}`)
      - traefik.http.routers.mq.entrypoints=websecure
      - traefik.http.routers.mq.tls.certresolver=le
      - traefik.http.services.mq.loadbalancer.server.port=15672

  minio:
    image: quay.io/minio/minio:latest
    container_name: solstice-minio
    restart: unless-stopped
    command: server /data --console-address ":9001"
    environment:
      MINIO_ROOT_USER: ${MINIO_ROOT_USER}
      MINIO_ROOT_PASSWORD: ${MINIO_ROOT_PASSWORD}
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:9000/minio/health/live"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    volumes:
      - minio-data:/data:Z
    networks:
      - core
    labels:
      - traefik.enable=true
      # MinIO Console at minio.svc.${DOMAIN}
      - traefik.http.routers.minio.rule=Host(`minio.svc.${DOMAIN}`)
      - traefik.http.routers.minio.entrypoints=websecure
      - traefik.http.routers.minio.tls.certresolver=le
      - traefik.http.services.minio.loadbalancer.server.port=9001
      # S3 API via TCP router on s3.svc.${DOMAIN}
      - traefik.tcp.routers.s3.rule=HostSNI(`s3.svc.${DOMAIN}`)
      - traefik.tcp.routers.s3.entrypoints=websecure
      - traefik.tcp.routers.s3.tls=true
      - traefik.tcp.routers.s3.tls.certresolver=le
      - traefik.tcp.services.s3.loadbalancer.server.port=9000

  minio-setup:
    image: quay.io/minio/mc:latest
    container_name: solstice-minio-setup
    depends_on:
      minio:
        condition: service_healthy
    entrypoint: ["/bin/sh", "-c"]
    command: >-
      "mc alias set local http://minio:9000 ${MINIO_ROOT_USER} ${MINIO_ROOT_PASSWORD} &&
      mc mb -p local/solstice-logs-staging || true &&
      mc mb -p local/solstice-logs-prod || true &&
      if [ -n \"${MINIO_BUCKET}\" ]; then mc mb -p local/${MINIO_BUCKET} || true; fi"
    networks:
      - core

  orchestrator:
    build:
      context: ../..
      dockerfile: deploy/images/orchestrator/Containerfile
      args:
        BIN: orchestrator
    image: local/solstice-orchestrator:latest
    container_name: solstice-orchestrator
    restart: unless-stopped
    environment:
      RUST_LOG: info
      DATABASE_URL: postgres://${POSTGRES_USER}:${POSTGRES_PASSWORD}@postgres:5432/solstice_${ENV}
      AMQP_URL: amqp://${RABBITMQ_DEFAULT_USER}:${RABBITMQ_DEFAULT_PASS}@rabbitmq:5672/solstice-${ENV}
      AMQP_EXCHANGE: solstice.jobs
      AMQP_QUEUE: solstice.jobs.v1
      AMQP_ROUTING_KEY: jobrequest.v1
      HTTP_ADDR: 0.0.0.0:8081
      # URL where logs-service is exposed (used for redirects)
      LOGS_BASE_URL: https://logs.${ENV}.${DOMAIN}
      # Paths inside the container to runner binaries that will be uploaded over SSH
      RUNNER_LINUX_PATH: /opt/solstice/runners/solstice-runner-linux
      RUNNER_ILLUMOS_PATH: /opt/solstice/runners/solstice-runner-illumos
      # Remote path on the VM where the runner will be uploaded and executed
      REMOTE_RUNNER_PATH: /usr/local/bin/solstice-runner
      # SSH connect timeout for reaching the VM (seconds)
      SSH_CONNECT_TIMEOUT_SECS: ${SSH_CONNECT_TIMEOUT_SECS:-300}
      # Libvirt configuration for Linux/KVM
      LIBVIRT_URI: ${LIBVIRT_URI:-qemu:///system}
      LIBVIRT_NETWORK: ${LIBVIRT_NETWORK:-default}
    depends_on:
      postgres:
        condition: service_healthy
      postgres-setup:
        condition: service_completed_successfully
      rabbitmq:
        condition: service_healthy
    # Host integrations: libvirt sockets/devices and config + image/work directories
    volumes:
      # Read-only mount of the image map config into the container path expected by the binary
      - ${ORCH_IMAGE_MAP_PATH:-../../examples/orchestrator-image-map.yaml}:/examples/orchestrator-image-map.yaml:ro,Z
      # Writable bind for images so the orchestrator can download/retain base images on the host
      - ${ORCH_IMAGES_DIR:-/var/lib/solstice/images}:/var/lib/solstice/images:Z
      # Writable bind for per-VM overlays and console logs (used by libvirt backend)
      - ${ORCH_WORK_DIR:-/var/lib/solstice-ci}:/var/lib/solstice-ci:Z
      # Read-only bind for locally built workflow runner binaries; orchestrator will upload over SSH
      - ${RUNNER_DIR_HOST:-../../target/runners}:/opt/solstice/runners:ro,Z
      # Libvirt control sockets (ro is sufficient for read-only, but write is needed to create domains)
      - /var/run/libvirt/libvirt-sock:/var/run/libvirt/libvirt-sock:Z
      - /var/run/libvirt/libvirt-sock-ro:/var/run/libvirt/libvirt-sock-ro:Z
      # Optional: expose host libvirt configs for network XML reads
      - /etc/libvirt:/etc/libvirt:ro,Z
      - /var/lib/libvirt:/var/lib/libvirt:ro,Z
      # KVM device for hardware acceleration
      - /dev/kvm:/dev/kvm
    devices:
      - /dev/kvm
    networks:
      - core
    labels:
      - traefik.enable=true
      # HTTP endpoints at api.${ENV}.${DOMAIN}
      - traefik.http.routers.api.rule=Host(`api.${ENV}.${DOMAIN}`)
      - traefik.http.routers.api.entrypoints=websecure
      - traefik.http.routers.api.tls.certresolver=le
      - traefik.http.services.api.loadbalancer.server.port=8081

  logs-service:
    build:
      context: ../..
      dockerfile: deploy/images/logs-service/Containerfile
    image: local/solstice-logs-service:latest
    container_name: solstice-logs-service
    restart: unless-stopped
    environment:
      RUST_LOG: info
      DATABASE_URL: postgres://${POSTGRES_USER}:${POSTGRES_PASSWORD}@postgres:5432/solstice_${ENV}
      HTTP_ADDR: 0.0.0.0:8082
    depends_on:
      postgres:
        condition: service_healthy
      postgres-setup:
        condition: service_completed_successfully
    networks:
      - core
    labels:
      - traefik.enable=true
      # Expose logs service at logs.${ENV}.${DOMAIN}
      - traefik.http.routers.logs.rule=Host(`logs.${ENV}.${DOMAIN}`)
      - traefik.http.routers.logs.entrypoints=websecure
      - traefik.http.routers.logs.tls.certresolver=le
      - traefik.http.services.logs.loadbalancer.server.port=8082

  forge-integration:
    build:
      context: ../..
      dockerfile: deploy/images/forge-integration/Containerfile
      args:
        BIN: forge-integration
    image: local/solstice-forge-integration:latest
    container_name: solstice-forge-integration
    restart: unless-stopped
    environment:
      RUST_LOG: info
      AMQP_URL: amqp://${RABBITMQ_DEFAULT_USER}:${RABBITMQ_DEFAULT_PASS}@rabbitmq:5672/solstice-${ENV}
      # HTTP server config for webhooks
      HTTP_ADDR: 0.0.0.0:8080
      WEBHOOK_PATH: /webhooks/forgejo
      # Secrets and Forgejo API configuration
      WEBHOOK_SECRET: ${WEBHOOK_SECRET}
      FORGEJO_TOKEN: ${FORGEJO_TOKEN}
      FORGEJO_BASE_URL: ${FORGEJO_BASE_URL}
      # URL where logs-service is exposed (used for commit status links)
      LOGS_BASE_URL: https://logs.${ENV}.${DOMAIN}
    depends_on:
      rabbitmq:
        condition: service_healthy
    networks:
      - core
    labels:
      - traefik.enable=true
      # Forge webhooks at forge.svc.${DOMAIN}
      - traefik.http.routers.forge.rule=Host(`forge.${ENV}.${DOMAIN}`)
      - traefik.http.routers.forge.entrypoints=websecure
      - traefik.http.routers.forge.tls.certresolver=le
      - traefik.http.services.forge.loadbalancer.server.port=8080

  github-integration:
    build:
      context: ../..
      dockerfile: deploy/images/github-integration/Containerfile
      args:
        BIN: github-integration
    image: local/solstice-github-integration:latest
    container_name: solstice-github-integration
    restart: unless-stopped
    environment:
      RUST_LOG: info
      AMQP_URL: amqp://${RABBITMQ_DEFAULT_USER}:${RABBITMQ_DEFAULT_PASS}@rabbitmq:5672/solstice-${ENV}
      # HTTP server for GitHub webhooks
      HTTP_ADDR: 0.0.0.0:8082
      WEBHOOK_PATH: /webhooks/github
      GITHUB_WEBHOOK_SECRET: ${GITHUB_WEBHOOK_SECRET}
      HOOKDECK_SIGNING_SECRET: ${HOOKDECK_SIGNING_SECRET}
      GITHUB_APP_ID: ${GITHUB_APP_ID}
      GITHUB_APP_KEY_PATH: /app/github-app-key.pem
      GITHUB_API_BASE: ${GITHUB_API_BASE:-https://api.github.com}
      GITHUB_CHECK_NAME: ${GITHUB_CHECK_NAME:-Solstice CI}
      # URL where logs-service is exposed (used for check-run links)
      LOGS_BASE_URL: https://logs.${ENV}.${DOMAIN}
    volumes:
      - ${GITHUB_APP_KEY_PATH}:/app/github-app-key.pem
    depends_on:
      rabbitmq:
        condition: service_healthy
    networks:
      - core
    labels:
      - traefik.enable=true
      # GitHub webhooks at github.svc.${DOMAIN}
      - traefik.http.routers.github.rule=Host(`github.${ENV}.${DOMAIN}`)
      - traefik.http.routers.github.entrypoints=websecure
      - traefik.http.routers.github.tls.certresolver=le
      - traefik.http.services.github.loadbalancer.server.port=8082

  runner-integration:
    build:
      context: ../..
      dockerfile: deploy/images/runner-integration/Containerfile
    image: local/solstice-runner-integration:latest
    container_name: solstice-runner-integration
    restart: unless-stopped
    environment:
      RUST_LOG: info
      AMQP_URL: amqp://${RABBITMQ_DEFAULT_USER}:${RABBITMQ_DEFAULT_PASS}@rabbitmq:5672/solstice-${ENV}
      AMQP_RESULTS_QUEUE: solstice.runner-results.v1
      # Forgejo runner configuration
      FORGEJO_URL: ${FORGEJO_URL}
      # API base for fetching workflow files from repos (self-hosted Forgejo)
      FORGEJO_BASE_URL: ${FORGEJO_URL}/api/v1
      FORGEJO_TOKEN: ${RUNNER_FORGEJO_TOKEN}
      RUNNER_REGISTRATION_TOKEN: ${RUNNER_REGISTRATION_TOKEN}
      RUNNER_NAME: solstice-runner-${ENV}
      RUNNER_LABELS: ${RUNNER_LABELS:-self-hosted}
      RUNNER_STATE_PATH: /data/runner-state.json
      MAX_CONCURRENCY: ${RUNNER_MAX_CONCURRENCY:-4}
      # Internal URL for fetching logs (same Docker network)
      LOGS_BASE_URL: http://solstice-logs-service:8082
    depends_on:
      rabbitmq:
        condition: service_healthy
      logs-service:
        condition: service_started
    volumes:
      - runner-state:/data:Z
    networks:
      - core