Docker-Compose无法同时启动两个Node.js应用与MongoDB容器问题

问题描述 #

我有两个位于独立文件夹中的Node.js应用，尝试通过docker-compose运行这两个应用与一个MongoDB容器。

完整的docker-compose配置如下：

version: "3.8"
services:

  service-a:
    container_name: service-a
    build: ./service-a
    restart: unless-stopped
    ports:
      - '3000:3000'
    stdin_open: true
    tty: true
    depends_on:
      - mongo

  service-b:
    container_name: service-b
    build: ./service-b
    restart: unless-stopped
    ports:
      - '3001:3001'
    stdin_open: true
    tty: true
    depends_on:
      - mongo

  mongo:
    container_name: mongo
    image: mongo:latest
    ports:
      - '27017:27017'

volumes:
  db:

执行docker-compose up时无法同时启动三个容器，但注释掉其中一个应用后（如下配置）就能正常运行：

version: "3.8"
services:

  service-b:
    container_name: service-b
    build: ./service-b
    restart: unless-stopped
    ports:
      - '3001:3001'
    stdin_open: true
    tty: true
    depends_on:
      - mongo

  mongo:
    container_name: mongo
    image: mongo:latest
    ports:
      - '27017:27017'

volumes:
  db:

需要解决的问题：如何让三个容器同时运行，且两个Node.js应用都能连接到MongoDB容器？

附docker-compose up日志：

{
  "t": {
    "$date": "2022-11-15T14:32:20.034+00:00"
  },
  "s": "I",
  "c": "COMMAND",
  "id": 20499,
  "ctx": "ftdc",
  "msg": "serverStatus was very slow",
  "attr": {
    "timeStats": {
      "after basic": 248,
      "after activeIndexBuilds": 364,
      "after asserts": 526,
      "after batchedDeletes": 582,
      "after bucketCatalog": 726,
      "after catalogStats": 1007,
      "after connections": 1143,
      "after electionMetrics": 1434,
      "after extra_info": 1627,
      "after flowControl": 1658,
      "after freeMonitoring": 1715,
      "after globalLock": 1715,
      "after indexBulkBuilder": 1735,
      "after indexStats": 1735,
      "after locks": 1735,
      "after logicalSessionRecordCache": 1750,
      "after mirroredReads": 1750,
      "after network": 1775,
      "after opLatencies": 1775,
      "after opcounters": 1785,
      "after opcountersRepl": 1785,
      "after oplog": 1785,
      "after oplogTruncation": 1795,
      "after readConcernCounters": 1818,
      "after repl": 1903,
      "after scramCache": 2002,
      "after security": 2090,
      "after storageEngine": 2150,
      "after tcmalloc": 2239,
      "after tenantMigrations": 2239,
      "after trafficRecording": 2239,
      "after transactions": 2251,
      "after transportSecurity": 2251,
      "after twoPhaseCommitCoordinator": 2251,
      "after wiredTiger": 2382,
      "at end": 2716
    }
  }
}

解决步骤 #

1. 确保MongoDB就绪后再启动应用 #

depends_on仅保证容器启动顺序，不确保MongoDB服务完全就绪。两个应用同时启动时，可能在MongoDB还未初始化完成就发起连接，导致应用容器崩溃。

修改方案：

• 给MongoDB添加健康检查，让应用等待其就绪后再启动：
  更新docker-compose中mongo服务的配置：

  mongo:
    container_name: mongo
    image: mongo:latest
    ports:
      - '27017:27017'
    healthcheck:
      test: ["CMD", "mongosh", "--eval", "db.adminCommand('ping')"]
      interval: 5s
      timeout: 5s
      retries: 5
  

  同时修改两个应用的depends_on配置：

  service-a:
    # 其他配置保留
    depends_on:
      mongo:
        condition: service_healthy
  service-b:
    # 其他配置保留
    depends_on:
      mongo:
        condition: service_healthy
  

• 在Node.js应用中添加连接重试逻辑（以mongoose为例）：

  const mongoose = require('mongoose');
  
  const connectWithRetry = () => {
    mongoose.connect('mongodb://mongo:27017/your-db-name')
      .then(() => console.log('MongoDB连接成功'))
      .catch(err => {
        console.log('MongoDB连接失败，5秒后重试...', err);
        setTimeout(connectWithRetry, 5000);
      });
  };
  
  connectWithRetry();
  

2. 限制容器资源占用 #

日志显示MongoDB的serverStatus执行缓慢，说明宿主机CPU/内存资源可能不足，同时启动三个容器导致资源耗尽。

修改方案：
给每个容器添加资源限制，避免单个容器占用过多资源：

service-a:
  # 其他配置保留
  deploy:
    resources:
      limits:
        cpus: '0.5'
        memory: 512M
      reservations:
        cpus: '0.25'
        memory: 256M
service-b:
  # 其他配置保留
  deploy:
    resources:
      limits:
        cpus: '0.5'
        memory: 512M
      reservations:
        cpus: '0.25'
        memory: 256M
mongo:
  # 其他配置保留
  deploy:
    resources:
      limits:
        cpus: '1'
        memory: 1G
      reservations:
        cpus: '0.5'
        memory: 512M

3. 检查应用Dockerfile与连接字符串 #

• 确认两个应用的Dockerfile正确配置，示例如下：

  FROM node:18-alpine
  WORKDIR /app
  COPY package*.json ./
  RUN npm install
  COPY . .
  CMD ["node", "server.js"]
  

• 应用中MongoDB连接字符串必须使用服务名mongo，不能用localhost：
  正确写法：mongodb://mongo:27017/your-db-name
  错误写法：mongodb://localhost:27017/your-db-name（容器内localhost指向自身，无法访问MongoDB容器）

4. 单独验证应用容器 #

分别单独构建并运行两个应用容器，确认它们能正常启动并连接MongoDB：

# 验证service-a
cd service-a
docker build -t service-a .
docker run --network=your-compose-network service-a

# 验证service-b同理
cd service-b
docker build -t service-b .
docker run --network=your-compose-network service-b

（your-compose-network可通过docker network ls查看，docker-compose默认会创建一个以项目名为前缀的网络）

内容的提问来源于stack exchange，提问作者Khaled_ch