docker-compose.dev.yml

services:
  traefik:
    image: traefik:latest
    restart: unless-stopped
    security_opt:
      - label:disable
    user: "0"
    ports:
      - "8080:80"
      - "8443:443"
      - "8081:8080"
    networks:
      - telemetry-network
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - ./traefik:/etc/traefik:ro
      - traefik-logs:/var/log/traefik
    environment:
      - TRAEFIK_DASHBOARD_AUTH=${TRAEFIK_DASHBOARD_AUTH}
      - LOCAL_DOMAIN=${LOCAL_DOMAIN:-localhost}
    healthcheck:
      test: ["CMD", "traefik", "healthcheck"]
      interval: 30s
      timeout: 10s
      retries: 3
    labels:
      - "traefik.enable=true"
      # Dashboard accessible via both domains with enhanced security
      - "traefik.http.routers.dashboard-local.rule=Host(`${LOCAL_DOMAIN:-localhost}`) && PathPrefix(`/net-dash`)"
      - "traefik.http.routers.dashboard-local.entrypoints=websecure"
      - "traefik.http.routers.dashboard-local.service=api@internal"
      - "traefik.http.routers.dashboard-local.tls=true"
      - "traefik.http.routers.dashboard-local.middlewares=dashboard-auth,secure-dashboard@file"
      - "traefik.http.middlewares.dashboard-auth.basicauth.users=${TRAEFIK_DASHBOARD_AUTH}"

  questdb:
    image: questdb/questdb:latest
    container_name: questdb
    restart: unless-stopped
    ports:
      - "9000:9000" 
      - "8812:8812" 
      - "9009:9009" 
      - "9003:9003" 
    volumes:
      - questdb-data:/var/lib/questdb
    environment:
      JAVA_OPTS: "-Xmx2g -Xms1g -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:G1ReservePercent=20"
      QDB_SHARED_WORKER_COUNT: "5"
      QDB_HTTP_ENABLED: "true"
      QDB_PG_ENABLED: "true"
      QDB_LINE_TCP_ENABLED: "true"
      QDB_METRICS_ENABLED: "true"
      QDB_CAIRO_WAL_ENABLED_DEFAULT: "true"
      QDB_CAIRO_COMMIT_LAG: "120000"
      QDB_CAIRO_MAX_UNCOMMITTED_ROWS: "50000"
    networks:
      - telemetry-network
    labels:
      - "traefik.enable=true"
      # QuestDB API routes (no web UI, direct API access)
      - "traefik.http.routers.questdb-local.rule=Host(`${LOCAL_DOMAIN}`) && PathPrefix(`/questdb`)"
      - "traefik.http.routers.questdb-local.entrypoints=websecure"
      - "traefik.http.routers.questdb-local.tls=true"
      - "traefik.http.routers.questdb-local.middlewares=questdb-stripprefix"
      - "traefik.http.middlewares.questdb-stripprefix.stripprefix.prefixes=/questdb"
      - "traefik.http.services.questdb.loadbalancer.server.port=9000"

  telemetry-service:
    build:
      context: ./telemetryService/golang
      dockerfile: Dockerfile
    container_name: telemetry-service
    restart: unless-stopped
    networks:
      - telemetry-network
    environment:
      QUESTDB_URL: questdb:8812;username=admin;password=quest
      QUESTDB_HOST: questdb
      QUESTDB_PORT: 9000
      GOMAXPROCS: "6"       # Limit Go scheduler
      GOGC: "200"           # Less aggressive GC
    ports:
      - "9092:9092"  # Prometheus metrics
      - "8010:8010"  # Prometheus metrics
    depends_on:
      questdb:
        condition: service_started
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.telemetry-api-local.rule=Host(`${LOCAL_DOMAIN:-localhost}`) &&PathPrefix(`/api`)"
      - "traefik.http.routers.telemetry-api-local.entrypoints=web"
      - "traefik.http.routers.telemetry-api-local.middlewares=api-security@file"
      - "traefik.http.services.telemetry-service.loadbalancer.server.port=8010"

  telemetry-dashboard:
    build:
      context: ./dashboard
      dockerfile: Dockerfile
    container_name: telemetry-dashboard
    restart: unless-stopped
    networks:
      - telemetry-network
    deploy:
      resources:
        limits:
          cpus: '1.0'      
          memory: 1g       
    labels:
      - "traefik.enable=true"
      # Dashboard main routes
      - "traefik.http.routers.telemetry-dashboard-local.rule=Host(`${LOCAL_DOMAIN:-localhost}`) &&PathPrefix(`/dashboard`)"
      - "traefik.http.routers.telemetry-dashboard-local.entrypoints=web"
      # Service configuration
      - "traefik.http.services.telemetry-dashboard.loadbalancer.server.port=8080"
      - "com.centurylinklabs.watchtower.enable=true"

  grafana:
    image: dhi.io/grafana:12.3
    container_name: grafana
    restart: unless-stopped
    volumes:
      - grafana-data:/var/lib/grafana
      - ./config/grafana:/etc/grafana/provisioning:ro
    ports:
      - "3002:3000"
    environment:
      # Grafana settings optimized for high performance
      GF_SECURITY_ADMIN_PASSWORD: ${GRAFANA_ADMIN_PASSWORD:-admin123}
      GF_USERS_ALLOW_SIGN_UP: "false"
      GF_SERVER_ROOT_URL: http://localhost:3002
      GF_SERVER_SERVE_FROM_SUB_PATH: "false"
      GF_INSTALL_PLUGINS: grafana-clock-panel
      GF_PROVISIONING_AUTO_ASSIGN_ORG: "true"
      GF_PROVISIONING_AUTO_ASSIGN_ORG_ROLE: Admin
      
      GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH: /etc/grafana/provisioning/dashboards/files/system-monitoring.json
      GF_USERS_DEFAULT_THEME: dark
      GF_ANALYTICS_REPORTING_ENABLED: "false"
      GF_ANALYTICS_CHECK_FOR_UPDATES: "false"
      
    networks:
      - telemetry-network
    deploy:
      resources:
        limits:
          cpus: '0.75'     
          memory: 1.5g     
        reservations:
          cpus: '0.25'    
          memory: 256m
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.grafana-local.rule=Host(`${LOCAL_DOMAIN:-localhost}`) &&PathPrefix(`/grafana`)"
      - "traefik.http.routers.grafana-local.entrypoints=websecure"
      - "traefik.http.routers.grafana-local.tls=true"
      - "traefik.http.routers.grafana-local.middlewares=secure-management@file"
      - "traefik.http.services.grafana.loadbalancer.server.port=3000"

  prometheus:
    image: prom/prometheus:v2.47.2
    container_name: prometheus
    restart: unless-stopped
    volumes:
      - ./config/prometheus.yml:/etc/prometheus/prometheus.yml:ro
      - prometheus-data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.path=/prometheus'
      - '--storage.tsdb.retention.time=30d'  # Extended retention with 16GB RAM
      - '--web.console.libraries=/etc/prometheus/console_libraries'
      - '--web.console.templates=/etc/prometheus/consoles'
      - '--web.enable-lifecycle'
      - '--web.external-url=https://${LOCAL_DOMAIN:-localhost}/prometheus'
      - '--web.route-prefix=/'
      - '--storage.tsdb.retention.size=4GB'   # Use more disk space for metrics
    networks:
      - telemetry-network
    deploy:
      resources:
        limits:
          cpus: '0.75'     # More CPU for metrics processing
          memory: 2g       # Double memory for larger datasets
        reservations:
          cpus: '0.25'
          memory: 512m
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.prometheus-local.rule=Host(`${LOCAL_DOMAIN:-localhost}`) && PathPrefix(`/prometheus`)"
      - "traefik.http.routers.prometheus-local.entrypoints=websecure"
      - "traefik.http.routers.prometheus-local.tls=true"
      - "traefik.http.routers.prometheus-local.middlewares=prometheus-stripprefix,secure-management@file"
      - "traefik.http.middlewares.prometheus-stripprefix.stripprefix.prefixes=/prometheus"
      - "traefik.http.services.prometheus.loadbalancer.server.port=9090"

  # Simplified monitoring stack
  node-exporter:
    image: prom/node-exporter:v1.6.1
    container_name: node-exporter
    restart: unless-stopped
    volumes:
      - /proc:/host/proc:ro
      - /sys:/host/sys:ro
      - /:/rootfs:ro
    command:
      - '--path.procfs=/host/proc'
      - '--path.rootfs=/rootfs'
      - '--path.sysfs=/host/sys'
      - '--collector.filesystem.mount-points-exclude=^/(sys|proc|dev|host|etc)($$|/)'
    networks:
      - telemetry-network
    deploy:
      resources:
        limits:
          cpus: 0.2
          memory: 128m

volumes:
  questdb-data:
    driver: local
  grafana-data:
    driver: local
  prometheus-data:
    driver: local
  traefik-logs:
    driver: local

networks:
  telemetry-network:
    driver: bridge