managarten/docker/prometheus/prometheus.yml

# ManaCore Prometheus Configuration
# Scrapes metrics from all services

global:
  scrape_interval: 15s
  evaluation_interval: 15s

# Load alerting rules
rule_files:
  - /etc/prometheus/alerts.yml

# Alertmanager configuration
alerting:
  alertmanagers:
    - static_configs:
        - targets: ['alertmanager:9093']

scrape_configs:
  # Prometheus self-monitoring
  - job_name: 'prometheus'
    static_configs:
      - targets: ['localhost:9090']

  # Host system metrics via node-exporter
  - job_name: 'node'
    static_configs:
      - targets: ['node-exporter:9100']
    relabel_configs:
      - source_labels: [__address__]
        target_label: instance
        replacement: 'mac-mini'

  # Docker container metrics via cAdvisor
  - job_name: 'cadvisor'
    static_configs:
      - targets: ['cadvisor:8080']

  # PostgreSQL metrics
  - job_name: 'postgres'
    static_configs:
      - targets: ['postgres-exporter:9187']

  # Redis metrics
  - job_name: 'redis'
    static_configs:
      - targets: ['redis-exporter:9121']

  # ============================================
  # Application Backends (after /metrics added)
  # ============================================

  # Auth Service
  - job_name: 'mana-core-auth'
    static_configs:
      - targets: ['mana-core-auth:3001']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Chat Backend
  - job_name: 'chat-backend'
    static_configs:
      - targets: ['chat-backend:3030']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Todo Backend
  - job_name: 'todo-backend'
    static_configs:
      - targets: ['todo-backend:3031']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Calendar Backend
  - job_name: 'calendar-backend'
    static_configs:
      - targets: ['calendar-backend:3032']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Clock Backend: REMOVED — migrated to local-first

  # Contacts Backend
  - job_name: 'contacts-backend'
    static_configs:
      - targets: ['contacts-backend:3034']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Storage Backend
  - job_name: 'storage-backend'
    static_configs:
      - targets: ['storage-backend:3035']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Presi Backend: REMOVED — replaced by Hono server

  # Nutriphi Backend
  - job_name: 'nutriphi-backend'
    static_configs:
      - targets: ['nutriphi-backend:3037']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # SkillTree Backend
  - job_name: 'skilltree-backend'
    static_configs:
      - targets: ['skilltree-backend:3038']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Photos Backend: REMOVED — migrated to local-first + direct mana-media

  # Zitare Backend: REMOVED — migrated to local-first

  # Mukke Backend
  - job_name: 'mukke-backend'
    static_configs:
      - targets: ['mukke-backend:3010']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Planta Backend
  - job_name: 'planta-backend'
    static_configs:
      - targets: ['planta-backend:3022']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Picture Backend
  - job_name: 'picture-backend'
    static_configs:
      - targets: ['picture-backend:3040']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # ============================================
  # Core Services
  # ============================================

  # Mana LLM Gateway (Ollama + Google Fallback)
  - job_name: 'mana-llm'
    static_configs:
      - targets: ['mana-llm:3025']
    metrics_path: '/metrics'
    scrape_interval: 15s

  # Mana Search Service
  - job_name: 'mana-search'
    static_configs:
      - targets: ['mana-search:3020']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Mana Media Service
  - job_name: 'mana-media'
    static_configs:
      - targets: ['mana-media:3015']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Matrix Synapse
  - job_name: 'synapse'
    static_configs:
      - targets: ['synapse:9002']
    metrics_path: '/_synapse/metrics'
    scrape_interval: 30s

  # ============================================
  # GPU Server (Windows PC, LAN: 192.168.178.11)
  # ============================================

  # GPU: LLM Gateway
  - job_name: 'gpu-llm'
    static_configs:
      - targets: ['192.168.178.11:3025']
        labels:
          instance: 'gpu-server'
    metrics_path: '/metrics'
    scrape_interval: 15s

  # GPU: Speech-to-Text (WhisperX)
  - job_name: 'gpu-stt'
    static_configs:
      - targets: ['192.168.178.11:3020']
        labels:
          instance: 'gpu-server'
    metrics_path: '/health'
    scrape_interval: 30s

  # GPU: Text-to-Speech
  - job_name: 'gpu-tts'
    static_configs:
      - targets: ['192.168.178.11:3022']
        labels:
          instance: 'gpu-server'
    metrics_path: '/health'
    scrape_interval: 30s

  # GPU: Image Generation (FLUX.2)
  - job_name: 'gpu-image-gen'
    static_configs:
      - targets: ['192.168.178.11:3023']
        labels:
          instance: 'gpu-server'
    metrics_path: '/health'
    scrape_interval: 30s

  # ============================================
  # Go Infrastructure Services
  # ============================================

  # API Gateway (Go)
  - job_name: 'mana-api-gateway'
    static_configs:
      - targets: ['mana-api-gateway:3060']
    metrics_path: '/metrics'
    scrape_interval: 15s

  # Matrix Bot (Go) — consolidated 21 bots
  - job_name: 'mana-matrix-bot'
    static_configs:
      - targets: ['mana-matrix-bot:4000']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Sync Server (Go) — local-first data sync
  - job_name: 'mana-sync'
    static_configs:
      - targets: ['mana-core-sync:3051']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Notification Service (Go) — email, push, matrix, webhook
  - job_name: 'mana-notify'
    static_configs:
      - targets: ['mana-core-notify:3042']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # Crawler Service (Go)
  - job_name: 'mana-crawler'
    static_configs:
      - targets: ['mana-crawler:3023']
    metrics_path: '/metrics'
    scrape_interval: 30s

  # ============================================
  # Pushgateway (deploy metrics, batch jobs)
  # ============================================
  - job_name: 'pushgateway'
    honor_labels: true
    static_configs:
      - targets: ['pushgateway:9091']