使用psycopg2连接Postgres容器时出现连接拒绝错误

部署环境 #

• postgres:10.4容器
• NestJS容器（服务名alyanetworkapi）
• 集成psycopg2的python:3.9.14-alpine3.16容器
  所有容器通过docker-compose.yml编排，NestJS可正常连接PostgreSQL，但Python容器连接失败。

docker-compose.yml 配置 #

version: "3"
services:
  alyanetworkapi:
    container_name: alyanetworkapi
    image: alyanetworkapi:1.0.0
    build:
      context: ./docker/alyanetworkapi
      args:
        DATABASE_USERNAME: ${DATABASE_USERNAME}
        DATABASE_DATABASE: ${DATABASE_DATABASE}
        DATABASE_HOST: ${DATABASE_HOST}
        DATABASE_PASSWORD: ${DATABASE_PASSWORD}
    ports:
      - "6002:8000"
    networks:
      - alyattes-backend
      - internal
    extra_hosts:
      - "host.docker.internal:host-gateway"
    depends_on:
      - postgres
    restart: unless-stopped
  postgres:
    image: postgres:10.4
    ports:
      - "5432:5432"
    environment:
      POSTGRES_USER: ${DATABASE_USERNAME}
      POSTGRES_PASSWORD: ${DATABASE_PASSWORD}
      POSTGRES_DB: ${DATABASE_DATABASE}
    user: postgres
    volumes:
      - ./database:/var/lib/postgresql/data
    networks:
      - alyattes-backend
      - internal
networks:
  alyattes-backend:
    driver: bridge
  internal:

Python容器Dockerfile #

FROM python:3.9.14-alpine3.16

WORKDIR /app

RUN apk update
RUN apk add git
RUN apk add --update alpine-sdk
RUN apk add libpq-dev

RUN apk update \
    && apk add --virtual build-deps gcc python3-dev musl-dev \
    && apk add postgresql-dev gcc python3-dev musl-dev \
    && apk del build-deps \
    && apk --no-cache add musl-dev linux-headers g++ bash

ARG DATABASE_USERNAME
ARG DATABASE_DATABASE
ARG DATABASE_HOST
ARG DATABASE_PASSWORD

ENV DATABASE_USERNAME=${DATABASE_USERNAME}
ENV DATABASE_DATABASE=${DATABASE_DATABASE}
ENV DATABASE_HOST=${DATABASE_HOST}
ENV DATABASE_PASSWORD=${DATABASE_PASSWORD}

RUN git clone https://${GITHUB_PAT}@github.com/owner/reponame.git

RUN chmod 775 -R /app

WORKDIR /app/alya-network-api

RUN pip3 install .

CMD [ "python3", "main.py"]

错误信息 #

psycopg2.OperationalError: connection to server at "postgres" (192.168.112.2), port 5432 failed: Connection refused
        Is the server running on that host and accepting TCP/IP connections?

排查与解决方法 #

1. 给Python容器添加docker-compose服务配置 #

当前docker-compose.yml未定义Python服务，需将其加入services列表，并配置与PostgreSQL相同的网络，否则容器间无法互通。示例配置：

python-app:
    build:
      context: ./path/to/python/dockerfile  # 替换为Python Dockerfile所在目录
      args:
        DATABASE_USERNAME: ${DATABASE_USERNAME}
        DATABASE_DATABASE: ${DATABASE_DATABASE}
        DATABASE_HOST: postgres  # 直接用PostgreSQL服务名作为主机名
        DATABASE_PASSWORD: ${DATABASE_PASSWORD}
        GITHUB_PAT: ${GITHUB_PAT}
    networks:
      - alyattes-backend
      - internal
    depends_on:
      - postgres
    restart: unless-stopped

2. 确保PostgreSQL服务完全就绪 #

depends_on仅保证容器启动顺序，不保证PostgreSQL服务可用。添加健康检查，确保数据库就绪后再启动Python容器：
修改PostgreSQL服务配置：

postgres:
    # ...其他原有配置
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${DATABASE_USERNAME} -d ${DATABASE_DATABASE}"]
      interval: 5s
      timeout: 5s
      retries: 5

同时更新Python服务的depends_on：

python-app:
    # ...其他配置
    depends_on:
      postgres:
        condition: service_healthy

3. 配置PostgreSQL监听所有网卡 #

PostgreSQL默认可能仅监听localhost，需修改配置允许容器网络连接，可通过启动命令设置：

postgres:
    # ...其他原有配置
    command: postgres -c listen_addresses='*'

4. 验证环境变量正确性 #

进入Python容器检查DATABASE_HOST是否为postgres：

docker exec -it <python容器名> env | grep DATABASE_HOST

如果值不正确，可直接在Python服务的environment中强制设置：

python-app:
    # ...其他配置
    environment:
      DATABASE_HOST: postgres

内容的提问来源于stack exchange，提问作者albert112