infra: Compose-Stack mit Monitoring, ohne LiteLLM (Task 5.2)

- deploy/compose/infra: Dienst litellm entfernt (D4), Router auf LITELLM_BASE_URL
  (appsrv08, Port noch offen), alle Images gepinnt, Grafana und pve-exporter
  ohne Profile, node-exporter für den LXC, Zustand über KNASTER_STATE_DIR
- Prometheus: GPU-VMs per file_sd (node, gpu), Proxmox-Knoten per IP, infra-node
- Grafana: Datasource und Dashboards provisioniert (grafana.com 10347, 1860,
  14574 sowie knaster-Übersicht)
- Alertmanager: n8n-Receiver nur vorbereitet (auskommentiert)
- Ansible: Rollen infra-stack und gpu-exporters, Playbooks monitoring.yml und
  monitoring-targets.yml, Skript monitoring-targets.py (tofu output → file_sd)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

deploy/compose/infra/docker-compose.yml

@@ -1,5 +1,10 @@
-# Rolle "infra" (LXC/VM auf einem i7-Node): Router, LiteLLM, Postgres, Monitoring.
-# Start: cp .env.example .env && docker compose up -d
+# 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
@@ -13,29 +18,15 @@ services:
       context: ../../..
       dockerfile: deploy/docker/router.Dockerfile
     image: knaster/router:${KNASTER_VERSION:-dev}
-    env_file: [{path: .env, required: false}]
+    env_file: [{path: "${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}
-      LITELLM_BASE_URL: http://litellm:4000
+      KNASTER_DRY_RUN: ${KNASTER_DRY_RUN:-true}
+      LITELLM_BASE_URL: ${LITELLM_BASE_URL:?LITELLM_BASE_URL fehlt (LiteLLM auf appsrv08, D4)}
     ports: ["8080:8080"]
     depends_on:
       postgres: {condition: service_healthy}
-      litellm: {condition: service_started}
-
-  litellm:
-    <<: *restart
-    image: ghcr.io/berriai/litellm:${LITELLM_TAG:-main-stable}   # vor Deployment auf feste Version pinnen
-    command: ["--config", "/app/config.yaml", "--port", "4000"]
-    env_file: [{path: .env, required: false}]
-    volumes:
-      - ../../litellm/config.yaml:/app/config.yaml:ro
-    ports: ["4000:4000"]
-    healthcheck:
-      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:4000/health/liveliness', timeout=4)"]
-      interval: 30s
-      timeout: 5s
-      retries: 3
 
   tts-piper:
     <<: *restart
@@ -53,7 +44,7 @@ services:
 
   postgres:
     <<: *restart
-    image: pgvector/pgvector:pg16
+    image: pgvector/pgvector:0.8.7-pg16-trixie
     environment:
       POSTGRES_USER: knaster
       POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:?POSTGRES_PASSWORD fehlt}
@@ -66,42 +57,72 @@ services:
       timeout: 5s
       retries: 5
 
+  # --- Monitoring ---------------------------------------------------------------------------------
   prometheus:
     <<: *restart
-    image: prom/prometheus:v2.53.0
+    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:v0.27.0
-    command: ["--config.file=/etc/alertmanager/alertmanager.yml"]
+    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
-      - ./secrets/router_token:/etc/alertmanager/router_token:ro
+      - ${KNASTER_STATE_DIR:-.}/secrets/router_token:/etc/alertmanager/router_token:ro
+      - alertmanager:/alertmanager
     ports: ["9093:9093"]
 
   grafana:
     <<: *restart
-    image: grafana/grafana:11.1.0
-    profiles: ["dashboards"]
+    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:${PVE_EXPORTER_TAG:-latest}   # vor Deployment pinnen
-    profiles: ["proxmox"]
+    image: prompve/prometheus-pve-exporter:3.10.1
+    # Token monitoring@pve!prom (PVEAuditor) als PVE_USER/PVE_TOKEN_NAME/PVE_TOKEN_VALUE/PVE_VERIFY_SSL
+    env_file: [{path: "${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:
-      - ./secrets/pve.yml:/etc/prometheus/pve.yml:ro
-    ports: ["9221:9221"]
+      - /:/host:ro   # ohne rslave: im unprivilegierten LXC nicht erlaubt
 
   autoheal:
     restart: unless-stopped
@@ -115,4 +136,5 @@ services:
 volumes:
   postgres: {}
   prometheus: {}
+  alertmanager: {}
   grafana: {}