cluster-knaster/deploy/compose/infra/docker-compose.yml

# Rolle "infra" (LXC 101 auf pve44, HA nach pve48/pve52): Router, TTS, Postgres und Monitoring in einem Stack.
# Kein eigener LiteLLM (D4): Der Router nutzt den LiteLLM mit PSE auf appsrv08 (LITELLM_BASE_URL).
#
# Lokal:   cp .env.example .env && docker compose up -d
# Cluster: Ansible-Rolle infra-stack (platform/proxmox/ansible). Zustand und Geheimnisse liegen dort
#          außerhalb des Repos in KNASTER_STATE_DIR=/opt/knaster/infra (.env, secrets/, file_sd/), Aufruf:
#          docker compose -p knaster-infra --env-file /opt/knaster/infra/.env -f <repo>/deploy/compose/infra/docker-compose.yml up -d
name: knaster-infra

x-restart: &restart
  restart: unless-stopped
  labels: {autoheal: "true"}

services:
  router:
    <<: *restart
    build:
      context: ../../..
      dockerfile: deploy/docker/router.Dockerfile
    image: knaster/router:${KNASTER_VERSION:-dev}
    env_file: [{path: "${KNASTER_STATE_DIR:-.}/.env", required: false}]
    environment:
      KNASTER_DB_URL: postgresql+psycopg://knaster:${POSTGRES_PASSWORD:?POSTGRES_PASSWORD fehlt}@postgres:5432/knaster
      KNASTER_SITE_DIR: /app/sites/${KNASTER_SITE:-knaster}
      KNASTER_DRY_RUN: ${KNASTER_DRY_RUN:-true}
      LITELLM_BASE_URL: ${LITELLM_BASE_URL:?LITELLM_BASE_URL fehlt (LiteLLM auf appsrv08, D4)}
    ports: ["8080:8080"]
    depends_on:
      postgres: {condition: service_healthy}

  tts-piper:
    <<: *restart
    build:
      context: ../../..
      dockerfile: deploy/docker/tts-piper.Dockerfile
    image: knaster/tts-piper:${KNASTER_VERSION:-dev}
    container_name: tts-piper
    environment:
      KNASTER_TTS_PARALLEL: ${TTS_PARALLEL:-2}   # i7 mit 4 Kernen: max. 2 parallele Synthesen
    ports: ["8090:8090"]
    deploy:
      resources:
        limits: {cpus: "2", memory: 1G}

  postgres:
    <<: *restart
    image: pgvector/pgvector:0.8.7-pg16-trixie
    environment:
      POSTGRES_USER: knaster
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:?POSTGRES_PASSWORD fehlt}
      POSTGRES_DB: knaster
    volumes:
      - postgres:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U knaster -d knaster"]
      interval: 10s
      timeout: 5s
      retries: 5

  # --- Monitoring ---------------------------------------------------------------------------------
  prometheus:
    <<: *restart
    image: prom/prometheus:v3.13.4
    command:
      - --config.file=/etc/prometheus/prometheus.yml
      - --storage.tsdb.path=/prometheus
      - --storage.tsdb.retention.time=15d
      - --web.enable-lifecycle
    volumes:
      - ../../prometheus:/etc/prometheus:ro
      # file_sd: Ziele der GPU-VMs (DHCP, IPs wechseln), erzeugt auf ops aus `tofu output -json gpu_vms`
      - ${KNASTER_STATE_DIR:-.}/file_sd:/etc/knaster-sd:ro
      - prometheus:/prometheus
    extra_hosts: ["host.docker.internal:host-gateway"]   # node-exporter des LXC läuft im Host-Netz
    ports: ["9090:9090"]

  alertmanager:
    <<: *restart
    image: prom/alertmanager:v0.34.1
    command:
      - --config.file=/etc/alertmanager/alertmanager.yml
      - --storage.path=/alertmanager
    volumes:
      - ../../alertmanager/alertmanager.yml:/etc/alertmanager/alertmanager.yml:ro
      - ${KNASTER_STATE_DIR:-.}/secrets/router_token:/etc/alertmanager/router_token:ro
      - alertmanager:/alertmanager
    ports: ["9093:9093"]

  grafana:
    <<: *restart
    image: grafana/grafana:12.4.12
    environment:
      GF_SECURITY_ADMIN_USER: admin
      GF_SECURITY_ADMIN_PASSWORD: ${GRAFANA_ADMIN_PASSWORD:?GRAFANA_ADMIN_PASSWORD fehlt}   # wirkt nur beim ersten Start
      GF_USERS_ALLOW_SIGN_UP: "false"
      GF_USERS_DEFAULT_THEME: dark
      GF_ANALYTICS_REPORTING_ENABLED: "false"
      GF_ANALYTICS_CHECK_FOR_UPDATES: "false"
      GF_ANALYTICS_CHECK_FOR_PLUGIN_UPDATES: "false"
      GF_NEWS_NEWS_FEED_ENABLED: "false"
      GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH: /var/lib/grafana/dashboards/knaster-uebersicht.json
    volumes:
      - ../../grafana/provisioning:/etc/grafana/provisioning:ro
      - ../../grafana/dashboards:/var/lib/grafana/dashboards:ro
      - grafana:/var/lib/grafana
    ports: ["3000:3000"]
    depends_on: [prometheus]

  pve-exporter:
    <<: *restart
    image: prompve/prometheus-pve-exporter:3.10.1
    # Token monitoring@pve!prom (PVEAuditor) als PVE_USER/PVE_TOKEN_NAME/PVE_TOKEN_VALUE/PVE_VERIFY_SSL
    env_file: [{path: "${KNASTER_STATE_DIR:-.}/secrets/pve-exporter.env", required: false}]
    expose: ["9221"]

  node-exporter:
    <<: *restart
    image: prom/node-exporter:v1.12.1
    command:
      - --path.rootfs=/host
      - --web.listen-address=:9100
    network_mode: host
    pid: host
    volumes:
      - /:/host:ro   # ohne rslave: im unprivilegierten LXC nicht erlaubt

  autoheal:
    restart: unless-stopped
    image: willfarrell/autoheal:1.2.0
    environment:
      AUTOHEAL_CONTAINER_LABEL: autoheal
      AUTOHEAL_INTERVAL: "15"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock

volumes:
  postgres: {}
  prometheus: {}
  alertmanager: {}
  grafana: {}