Docker环境下Prometheus无法访问Actuator端点问题求助

在Docker环境部署Prometheus容器后，无法从其他服务连接到/actuator/prometheus端点，出现两种错误：

Get "http://notification-service:8080/actuator/prometheus": dial tcp 172.24.0.13:8080: connect: connection refused

或者：

server returned HTTP status 404

Prometheus监控截图

另外，直接用容器IP发起curl请求时，初始返回404错误，但几分钟后能正常响应。

相关配置示例 #

docker-compose.yml 片段

services:
  ticket-service:
    container_name: ticket-service
    image: <docker_hub_nick>/ticket-service:latest
    environment:
      - SPRING_PROFILES_ACTIVE=docker
      - SPRING_DATASOURCE_URL=jdbc:postgresql://postgres-ticket:5431/ticket_service
    depends_on:
      - postgres-ticket
      - broker
      - zipkin
      - discovery-server
      - api-gateway

  prometheus:
    image: prom/prometheus:latest
    container_name: prometheus
    restart: always
    ports:
      - "9090:9090"
    volumes:
      - ./prometheus/prometheus.yml:/etc/prometheus/prometheus.yml
    depends_on:
      - bet-service
      - ticket-service
      - chat-service
      - notification-service
      - wallet-service

ticket-service 的 properties 配置

spring.application.name=ticket-service

management.health.circuitbreakers.enabled=true
management.endpoints.web.exposure.include=*
management.endpoint.health.show-details=always

prometheus.yml 配置

global:
  scrape_interval:     1m
  evaluation_interval: 30s
  scrape_timeout: 1m
  external_labels:
    monitor: 'easywin-monitor'

scrape_configs:
  - job_name: 'ticket_service'
    metrics_path: '/actuator/prometheus'
    static_configs:
      - targets: ['ticket-service:8080']
        labels:
          application: 'Ticket Service'

1. 修复服务启动顺序问题 #

depends_on仅保证容器启动顺序，不确保依赖服务完全就绪。Prometheus可能在Spring Boot服务初始化完成前就开始抓取，导致连接拒绝或404。

• 给Prometheus添加启动延迟脚本，修改command：

  prometheus:
    # ...其他配置
    command: ["sh", "-c", "sleep 60 && /bin/prometheus --config.file=/etc/prometheus/prometheus.yml"]
  

• 或者使用wait-for-it工具检测依赖服务端点就绪后再启动Prometheus

2. 验证Spring Boot端点暴露配置 #

• 确认docker环境对应的配置文件（如application-docker.properties）中包含完整的端点配置：

  management.endpoints.web.exposure.include=*
  management.endpoint.prometheus.enabled=true
  

• 检查服务启动日志，确认/actuator/prometheus端点已注册，搜索类似日志：Exposing X endpoint(s) beneath base path '/actuator'，需包含prometheus

3. 检查Docker网络连通性 #

• 进入Prometheus容器，执行ping ticket-service确认DNS解析正常
• 执行curl http://ticket-service:8080/actuator查看所有可用端点，确认prometheus存在
• 若DNS解析异常，为所有服务指定同一自定义网络：

  networks:
    app-network:
      driver: bridge
  
  services:
    ticket-service:
      # ...其他配置
      networks:
        - app-network
    prometheus:
      # ...其他配置
      networks:
        - app-network
  

4. 调整Prometheus抓取配置 #

• 缩短scrape_timeout（如改为10s），避免因服务初始化慢导致超时
• 添加重试和健康检查配置，仅当服务就绪后再抓取：

  scrape_configs:
    - job_name: 'ticket_service'
      metrics_path: '/actuator/prometheus'
      scrape_timeout: 10s
      static_configs:
        - targets: ['ticket-service:8080']
          labels:
            application: 'Ticket Service'
      retry_config:
        max_retries: 3
        retry_interval: 30s
      health_checks:
        - http_get:
            path: '/actuator/health'
            port: 8080
          interval: 30s
          timeout: 10s
  

5. 检查Spring Boot服务启动状态 #

• 查看ticket-service启动日志，确认数据库、注册中心等依赖服务已完全就绪，避免服务处于"启动中"状态时端点未激活

内容的提问来源于stack exchange，提问作者dr34mer