Docker生产环境下Django数据库无法更新的问题求助

问题现状 #

生产环境使用Django和Postgres时，数据库结构无法更新：

• 手动执行迁移命令，始终返回migrations are already applied或no changes are detected
• 本地migrations文件夹有新增迁移文件，但服务器上的迁移未实际生效
• 打开Django Admin面板出现数据库字段不存在的错误
• 目前只能通过docker compose down -v删除数据库重新初始化，导致数据丢失

核心原因 #

你的判断正确：Git忽略了migrations目录下的实际迁移文件，仅推送了migrations/__init__.py。这引发两个关键问题：

1. 容器构建时未包含本地生成的迁移文件，Django无法识别需要执行的迁移任务
2. entrypoint脚本在生产环境自动执行makemigrations，生成的临时迁移与本地开发的迁移不一致，而django_migrations表已记录旧迁移状态，最终导致模型与数据库结构不匹配

可行解决办法 #

1. 修正Git忽略规则，提交迁移文件到版本库 #

不要忽略migrations下的.py迁移文件，仅忽略缓存文件。修改.gitignore：

# 保留migrations目录下的所有.py文件，只忽略缓存
*/migrations/__pycache__/
*/migrations/*.pyc
!*/migrations/__init__.py
!*/migrations/*.py

提交所有本地已生成的迁移文件到Git，确保构建镜像时这些文件被包含。

2. 调整生产环境entrypoint脚本 #

生产环境禁止自动执行makemigrations，迁移文件必须在本地开发时生成并提交，生产仅执行migrate。修改entrypoint.sh：

#!/bin/sh

echo "----------python migrate----------"
python manage.py migrate --noinput

echo "----------python collectstatic----------"
python manage.py collectstatic --noinput

echo "----------python runserver----------"
uvicorn config.asgi:application --host 0.0.0.0 --port 8000 & python manage.py rqworker default --with-scheduler

exec "$@"

3. 修复当前生产数据库（无需删库） #

步骤1：备份数据库

先做全量备份，防止操作失误：

docker exec my_project-postgres pg_dump -U ${POSTGRES_USER} ${POSTGRES_DB} > production_backup.sql

步骤2：同步迁移文件到服务器

• 确保本地迁移文件已提交到Git，重新构建并推送镜像
• 拉取新镜像后重启容器，此时容器内已有正确的迁移文件

步骤3：执行迁移并修正状态

进入web容器，查看当前迁移状态：

docker exec -it ${API_CONTAINER_NAME} python manage.py showmigrations

如果有未应用的迁移，执行：

docker exec -it ${API_CONTAINER_NAME} python manage.py migrate --noinput

如果存在错误的迁移记录（比如之前自动生成的临时迁移），可针对指定应用重置后重新执行：

# 标记app_name的所有迁移为未应用
docker exec -it ${API_CONTAINER_NAME} python manage.py migrate --fake app_name zero
# 重新执行迁移
docker exec -it ${API_CONTAINER_NAME} python manage.py migrate app_name

附用户提供的配置文件 #

Dockerfile #

FROM python:3.10-buster

RUN apt-get update

#RUN useradd -m myapp

RUN mkdir -p /home/myapp/app
WORKDIR /home/myapp/app

ENV PYTHONDONTWRITEBYTECODE=1
ENV PYTHONUNBUFFERED=1
ENV PATH $PATH:/home/myapp/.local/bin

COPY requirements.txt .
RUN pip install --upgrade pip
RUN pip install -r requirements.txt

COPY . .

RUN chmod +x ./entrypoint.sh
ENTRYPOINT ["sh", "./entrypoint.sh"]

docker-compose.yml #

version: "3.8"

networks:
  my_project_network:
    name: my_project_network

services:
  web:
    image: ${TAG_LATEST}
    container_name: ${API_CONTAINER_NAME}
    command: ["./entrypoint.sh"]
    volumes:
      - my_project_media_data:/home/myapp/app/media/
      - my_project_static_data:/home/myapp/app/static/
    restart: always
    ports:
      - 8000:8000
    networks:
      - my_project_network
    depends_on:
      - db

  db:
    image: postgres:latest
    container_name: my_project-postgres
    volumes:
      - my_project_postgres_data:/var/lib/postgresql/data/
    environment:
      - POSTGRES_DB=${POSTGRES_DB}
      - POSTGRES_USER=${POSTGRES_USER}
      - POSTGRES_PASSWORD=${POSTGRES_PASSWORD}
      - POSTGRES_PORT=${POSTGRES_PORT}
    networks:
      - my_project_network
    restart: always

  redis:
    image: redis:latest
    container_name: my_project-redis
    ports:
      - 6379:6379
    networks:
      - my_project_network
    restart: always
    depends_on:
      - db

volumes:
  my_project_media_data:
    name: my_project_media_data
  my_project_static_data:
    name: my_project_static_data
  my_project_postgres_data:
    name: my_project_postgres_data

entrypoint.sh（原版本） #

#!/bin/sh

echo "----------python makemigrations----------"
python manage.py makemigrations --noinput
python manage.py migrate --noinput

echo "----------python collectstatic----------"
python manage.py collectstatic --noinput

echo "----------python runserver----------"
uvicorn config.asgi:application --host 0.0.0.0 --port 8000 & python manage.py rqworker default --with-scheduler

exec "$@"

gitlab-ci.yml #

image: alpine:3

stages:
  - build
  - deploy

variables:
  TAG_LATEST: $CI_REGISTRY_IMAGE/$CI_COMMIT_REF_NAME:latest
  TAG_COMMIT: $CI_REGISTRY_IMAGE/$CI_COMMIT_REF_NAME:$CI_COMMIT_SHORT_SHA

build:
  image: docker:latest
  stage: build
  only:
    - main
  services:
    - name: docker:dind
      alias: thedockerhost
  variables:
    DOCKER_HOST: tcp://thedockerhost:2375/
    DOCKER_DRIVER: overlay2
    DOCKER_TLS_CERTDIR: ""
  before_script:
    - chmod +x deploy.sh
    - ./deploy.sh
  script:
    - docker build -t $TAG_COMMIT -t $TAG_LATEST .
    - docker login -u $CI_REGISTRY_USER -p $CI_REGISTRY_PASSWORD $CI_REGISTRY
    - docker push $TAG_COMMIT
    - docker push $TAG_LATEST

deploy:
  stage: deploy
  only:
    - main
  before_script:
    - chmod +x deploy.sh
    - ./deploy.sh
  script:
    - chmod og= $ID_RSA
    - apk update && apk add openssh-client
    - ssh -i $ID_RSA -o StrictHostKeyChecking=no $SERVER_USER@$SERVER_IP -p $SERVER_SSH_PORT "sudo docker login -u $CI_REGISTRY_USER -p $CI_REGISTRY_PASSWORD $CI_REGISTRY"
    - ssh -i $ID_RSA -o StrictHostKeyChecking=no $SERVER_USER@$SERVER_IP -p $SERVER_SSH_PORT "sudo docker pull  $TAG_LATEST"
    - ssh -i $ID_RSA -o StrictHostKeyChecking=no $SERVER_USER@$SERVER_IP -p $SERVER_SSH_PORT "sudo docker container rm $API_CONTAINER_NAME -f || true"
    - scp -i $ID_RSA -o StrictHostKeyChecking=no -P $SERVER_SSH_PORT -r ./.env ./docker-compose.yml $SERVER_USER@$SERVER_IP:/home/dev/projects/my_project-service
    - ssh -i $ID_RSA -o StrictHostKeyChecking=no $SERVER_USER@$SERVER_IP -p $SERVER_SSH_PORT "cd /home/dev/projects/my_project-service && sudo docker compose up -d"

数据库错误截图 #

内容的提问来源于stack exchange，提问作者Fateme Fouladkar