使用AWS Copilot部署NestJS微服务至ECS时的TCP连接问题

本地运行正常的NestJS微服务栈（API Gateway + restaurant微服务），通过AWS Copilot部署到同一应用后，API Gateway无法连接restaurant服务，报错Error: connect ECONNREFUSED 0.0.0.0:8001。

环境与配置 #

服务类型 #

• API Gateway：Load Balanced Web Service
• restaurant服务：Backend Service
• 已在两者manifest中启用network.connect=true（Service Connect）

核心代码 #

Gateway/src/main.ts

import { NestFactory } from '@nestjs/core';
import { AppModule } from './app.module';

async function bootstrap() {
  const app = await NestFactory.create(AppModule);
  await app.listen(8000);
}
bootstrap();

Restaurant/src/main.ts

import { NestFactory } from '@nestjs/core';
import { AppModule } from './app.module';
import { Transport, TcpOptions } from '@nestjs/microservices';
async function bootstrap() {
  const app = await NestFactory.createMicroservice(AppModule, {
    transport: Transport.TCP,
    options: {
      host: '0.0.0.0',
      port: 8001,
    },
  } as TcpOptions);
  await app.listen();
}
bootstrap();

Gateway/src/app.module.ts

@Module({
  imports: [
    ConfigModule.forRoot({
      envFilePath: [`.env.stage.${process.env.STAGE}`],
    }),
  ],
  controllers: [RestaurantController, AppController],
  providers: [
    ConfigService,
    {
      provide: 'RESTAURANT_SERVICE',
      useFactory: (configService: ConfigService) => {
        return ClientProxyFactory.create({
          options: {
            host: '0.0.0.0',
            port: 8001,
          },
        });
      },
      inject: [ConfigService],
    },
  ],
})
export class AppModule {}

Manifest配置 #

api-gw manifest

name: api-gw
type: Load Balanced Web Service

http:
  path: '/'

image:
  build: Dockerfile
  port: 8000

cpu: 256       # Number of CPU units for the task.
memory: 512    # Amount of memory in MiB used by the task.
platform: linux/x86_64
count: 1       # Number of tasks that should be running in your service.
exec: true     # Enable running commands in your container.
network:
  connect: true # Enable Service Connect for intra-environment traffic between services.

restaurant service manifest

name: restaurant
type: Backend Service

image:
  build: Dockerfile
  port: 8001

cpu: 256       # Number of CPU units for the task.
memory: 512    # Amount of memory in MiB used by the task.
platform: linux/x86_64
count: 1       # Number of tasks that should be running in your service.
exec: true     # Enable running commands in your container.
network:
  connect: true # Enable Service Connect for intra-environment traffic between services.

核心问题 #

API Gateway中配置的restaurant服务地址0.0.0.0:8001是本地回环地址，在AWS Copilot的Service Connect网络环境中无法正确指向restaurant服务实例。

修复步骤 #

1. 修改API Gateway的服务连接配置
   在Gateway的app.module.ts中，将ClientProxy的host改为Copilot自动注入的服务DNS名称，同时补全传输类型配置：

   {
     provide: 'RESTAURANT_SERVICE',
     useFactory: (configService: ConfigService) => {
       return ClientProxyFactory.create({
         transport: Transport.TCP,
         options: {
           host: 'restaurant', // 使用manifest中定义的服务名称作为域名
           port: 8001,
         },
       });
     },
     inject: [ConfigService],
   },
   

   注：Copilot启用Service Connect后，同一应用内的服务可直接通过服务名称（manifest的name字段）作为域名访问，无需手动指定IP。

2. 验证端口一致性
   确保restaurant服务代码监听的端口（8001）与manifest中image.port配置完全一致，Copilot会自动通过Service Connect暴露该端口给同应用内的其他服务。

3. 重新部署服务
   执行部署命令使配置生效：

   copilot deploy --name api-gw
   copilot deploy --name restaurant
   

额外优化建议 #

通过环境变量管理服务地址，避免硬编码：

• 在Gateway的.env.stage.<环境名>文件中添加：

  RESTAURANT_SERVICE_HOST=restaurant
  RESTAURANT_SERVICE_PORT=8001
  

• 在app.module.ts中读取环境变量：

  options: {
    host: configService.get('RESTAURANT_SERVICE_HOST'),
    port: configService.get('RESTAURANT_SERVICE_PORT'),
  },
  

内容的提问来源于stack exchange，提问作者Mickael Zana