# pyMotor Observability Stack
# ----------------------------
# A self-contained Docker Compose deployment that mirrors NVIDIA Dynamo's
# local observability stack (Prometheus + Grafana + Tempo + Loki + OTel
# Collector + Exporters) for pyMotor.
#
# Profiles:
#   - default(no profile flag) : core stack
#   - npu : enable Ascend npu-exporter (requires Ascend drivers on host)
#
# Usage:
#   docker compose up -d
#   docker compose --profile npu up -d

name: pymotor-observability

x-image-prefix: &registry "${REGISTRY_PREFIX:-}"

networks:
  obs:
    driver: bridge

volumes:
  prometheus-data:
  grafana-data:
  tempo-data:
  loki-data:

services:
  # ---------------------------------------------------------------
  # Core: Prometheus
  # ---------------------------------------------------------------
  prometheus:
    image: ${REGISTRY_PREFIX:-}prom/prometheus:${PROMETHEUS_VERSION:-v2.55.1}
    pull_policy: if_not_present
    container_name: pymotor-prometheus
    restart: unless-stopped
    networks: [obs]
    ports:
      - "${PROMETHEUS_PORT:-9090}:9090"
    command:
      - --config.file=/etc/prometheus/prometheus.yml
      - --storage.tsdb.path=/prometheus
      - --storage.tsdb.retention.time=72h
      - --web.enable-lifecycle
      - --web.enable-remote-write-receiver
      # 保留 vllm:* 等带冒号的指标名（与 prometheus.yml 中
      # metric_name_validation_scheme: utf8 配套）
      - --enable-feature=utf8-names
    volumes:
      - ${PROMETHEUS_CONFIG_FILE:-./prometheus/prometheus.yml}:/etc/prometheus/prometheus.yml:ro
      - prometheus-data:/prometheus
    extra_hosts:
      - "host.docker.internal:host-gateway"

  # ---------------------------------------------------------------
  # Core: Tempo (traces)
  # ---------------------------------------------------------------
  tempo:
    image: ${REGISTRY_PREFIX:-}grafana/tempo:${TEMPO_VERSION:-2.6.1}
    pull_policy: if_not_present
    container_name: pymotor-tempo
    restart: unless-stopped
    networks: [obs]
    ports:
      - "${TEMPO_QUERY_PORT:-3200}:3200"
    command: ["-config.file=/etc/tempo/tempo.yaml"]
    volumes:
      - ./tempo/tempo.yaml:/etc/tempo/tempo.yaml:ro
      - tempo-data:/var/tempo

  # ---------------------------------------------------------------
  # Core: Loki (logs)
  # ---------------------------------------------------------------
  loki:
    image: ${LOKI_IMAGE:-grafana/loki:3.3.0}
    pull_policy: if_not_present
    container_name: pymotor-loki
    restart: unless-stopped
    networks: [obs]
    ports:
      - "${LOKI_PORT:-3100}:3100"
    command: ["-config.file=/etc/loki/loki.yaml"]
    volumes:
      - ./loki/loki.yaml:/etc/loki/loki.yaml:ro
      - loki-data:/var/loki

  # ---------------------------------------------------------------
  # Core: OpenTelemetry Collector
  # ---------------------------------------------------------------
  otel-collector:
    image: ${REGISTRY_PREFIX:-}otel/opentelemetry-collector-contrib:${OTEL_COLLECTOR_VERSION:-0.115.1}
    pull_policy: if_not_present
    container_name: pymotor-otel-collector
    restart: unless-stopped
    networks: [obs]
    depends_on: [tempo, loki]
    command: ["--config=/etc/otelcol/otel-collector.yaml"]
    volumes:
      - ${OTEL_CONFIG_FILE:-./otel-collector/otel-collector.yaml}:/etc/otelcol/otel-collector.yaml:ro
    ports:
      - "${OTEL_GRPC_PORT:-4317}:4317"
      - "${OTEL_HTTP_PORT:-4318}:4318"

  # ---------------------------------------------------------------
  # Core: Grafana (visualization)
  # ---------------------------------------------------------------
  grafana:
    # Use upstream image + volume-mounted provisioning (no local build / buildx).
    image: ${REGISTRY_PREFIX:-}grafana/grafana:${GRAFANA_VERSION:-11.3.0}
    pull_policy: if_not_present
    container_name: pymotor-grafana
    restart: unless-stopped
    networks: [obs]
    depends_on: [prometheus, tempo, loki]
    environment:
      GF_SECURITY_ADMIN_USER: ${GF_SECURITY_ADMIN_USER:-motor}
      GF_SECURITY_ADMIN_PASSWORD: ${GF_SECURITY_ADMIN_PASSWORD:-motor}
      GF_USERS_ALLOW_SIGN_UP: "false"
      GF_LOG_LEVEL: warn
      # Host proxy must not apply to in-compose datasources (prometheus/tempo).
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      http_proxy: ""
      https_proxy: ""
      NO_PROXY: prometheus,tempo,loki,otel-collector,localhost,127.0.0.1,host.docker.internal,.local,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16
      no_proxy: prometheus,tempo,loki,otel-collector,localhost,127.0.0.1,host.docker.internal,.local,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16
    ports:
      - "${GRAFANA_PORT:-3000}:3000"
    volumes:
      - grafana-data:/var/lib/grafana
      # Dashboards are baked into the image, but mount the source dir read-only
      # so the Grafana provisioner picks up live edits during development.
      - ./grafana/dashboards:/var/lib/grafana/dashboards:ro
      - ${GRAFANA_PROVISIONING_DIR:-./grafana/provisioning}:/etc/grafana/provisioning:ro

  # ---------------------------------------------------------------
  # Infra exporters
  # ---------------------------------------------------------------
  node-exporter:
    image: ${REGISTRY_PREFIX:-}prom/node-exporter:${NODE_EXPORTER_VERSION:-v1.8.2}
    container_name: pymotor-node-exporter
    restart: unless-stopped
    profiles: [full]
    networks: [obs]
    pid: host
    command:
      - --path.rootfs=/host
      - --collector.filesystem.mount-points-exclude=^/(sys|proc|dev|host|etc|var/lib/docker)($$|/)
    volumes:
      - /:/host:ro,rslave
    ports:
      - "${NODE_EXPORTER_PORT:-9100}:9100"

  cadvisor:
    image: ${REGISTRY_PREFIX:-}gcr.io/cadvisor/cadvisor:${CADVISOR_VERSION:-v0.49.1}
    container_name: pymotor-cadvisor
    restart: unless-stopped
    profiles: [full]
    networks: [obs]
    privileged: true
    devices:
      - /dev/kmsg
    volumes:
      - /:/rootfs:ro
      - /var/run:/var/run:ro
      - /sys:/sys:ro
      - /var/lib/docker/:/var/lib/docker:ro
    ports:
      - "${CADVISOR_PORT:-8088}:8080"
  # ---------------------------------------------------------------
  # Ascend NPU exporter (profile: npu) — requires host drivers
  # ---------------------------------------------------------------
  ascend-npu-exporter:
    image: ${NPU_EXPORTER_IMAGE:-swr.cn-south-1.myhuaweicloud.com/ascendhub/npu-exporter:v6.0.0}
    container_name: pymotor-npu-exporter
    restart: unless-stopped
    profiles: [npu]
    # NPU exporter needs host network + privileged + DCMI / driver mounts.
    network_mode: host
    privileged: true
    command:
      - --listen=0.0.0.0:${NPU_EXPORTER_PORT:-8082}
      - --updateTime=5
    volumes:
      - /usr/local/Ascend/driver:/usr/local/Ascend/driver:ro
      - /usr/local/dcmi:/usr/local/dcmi:ro
      - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi:ro
      - /var/log/Ascend:/var/log/Ascend
      - /etc/localtime:/etc/localtime:ro