How do you perform Django database migrations when using Docker-Compose? [closed]

I use this method:

services:
  web:
    build: .
    image: uzman
    command: python manage.py runserver 0.0.0.0:8000
    ports:
      - "3000:3000"
      - "8000:8000"
    volumes:
      - .:/code
    depends_on:
      - migration
      - db
  migration:
    image: uzman
    command: python manage.py migrate --noinput
    volumes:
      - .:/code
    depends_on:
      - db

Using docker hierarchy we made, the service migration runs after setting up the database and before running the main service. Now when you run your service docker will run migrations before running the server; look that the migration server is applied over the same image as the web server, it means that all migrations will be taken from your project, avoiding problems.

You avoid making entry points or whatever other thing this way.

Have your stack running then fire off a one shot docker-compose run command. E.g

#assume django in container named web
docker-compose run web python3 manage.py migrate

This works great for the built-in (default) SQLite database, but also for an external dockerized database that's listed as dependency. Here's an example docker-compose.yaml file

version: '3'

services:
  db:
    image: postgres
  web:
    build: .
    command: python manage.py runserver 0.0.0.0:8000
    volumes:
      - .:/code
    ports:
      - "8000:8000"
    depends_on:
      - db

https://docs.docker.com/compose/reference/run/

you can use docker-entrypoint.sh or a newer solution would be multiple comments in your docker-compose.yml

version: '3.7'

services:
  web:
    build: ./
    command: >
      sh -c "python manage.py collectstatic --noinput &&
             python manage.py migrate &&
             python manage.py runserver 0.0.0.0:8000"
    volumes:
      - ./:/usr/src/app/
    ports:
      - 8000:8000
    env_file:
      - ./.env
    depends_on:
      - postgres

  postgres:
    image: postgres:13.0-alpine
    ports:
      - 5432:5432
    volumes:
      - ./data/db:/var/lib/postgresql/data
    environment:
      - POSTGRES_USER=postgres
      - POSTGRES_PASSWORD=postgres
      - POSTGRES_DB=postgres

You can use docker exec command

docker exec -it container_id python manage.py migrate

If you have something like this in your docker-compose.yml

version: "3.7"

services:

  app:
    build:
      context: .
      dockerfile: docker/app/Dockerfile
    ports:
    - 8000:8000
    volumes:
        - ./:/usr/src/app
    depends_on:
      - db

  db:
    image: postgres
    restart: always
    environment:
      POSTGRES_USER: docker
      POSTGRES_PASSWORD: docker
      POSTGRES_DB: docker

Then you can simple run...

~$ docker-compose exec app python manage.py makemigrations
~$ docker-compose exec app python manage.py migrate

A common thing people get wrong with Docker + Django is when the migration commands should be run. Some common guidelines we should follow are:

• Don't run makemigrations and / or migrate in the Dockerfile. makemigrations is a command that is part of the development workflow, running it as part of the image creation can cause issues where you end up having an inconsistent history of migration files across images. migrate is part of the deployment, an image build step is not necessarily tied with a deployment and should hence be avoided.

• Don't run makemigrations as part of the entrypoint / command of your container. Containers are meant to be ephemeral, unless your files are mounted on a volume any changes you make in a container are lost when the container exits. This can cause you to have an inconsistent migration history.

• Avoid running migrate as part of the container that starts your server. When using Docker Compose or some other orchestrator, you might have multiple replicas of this container. Although Django does run the migrations in a transaction, its better not to take any chances.

Here is what we should be doing when containerizing a Django application:

1. Run makemigrations manually on your codebase as and when you change the models. This needs to be done out of the container (unless you're using something like devcontainers, or are mounting your code as a volume).
2. Have a container / docker compose service to run your Django application.
3. Have another container meant to run migrate, make sure the application containers depend on this and run after this has finished running. If using Kubernetes or similar try to set this up as a Job that can be run manually when needed.

Using docker exec, I was getting the following error:

AppRegistryNotReady("Models aren't loaded yet.")

So I used this command instead:

docker-compose -f local.yml run django python manage.py makemigrations

I know this is old, and maybe I am missing something here (if so, please enlighten me!), but why not just add the commands to your start.sh script, run by Docker to fire up your instance? It will take only a few extra seconds.

N.B. I set the DJANGO_SETTINGS_MODULE variable to make sure the correct database is used, as I use different databases for development and production (although I know this is not 'best practice').

This solved it for me:

#!/bin/bash
# Migrate the database first
echo "Migrating the database before starting the server"
export DJANGO_SETTINGS_MODULE="edatool.settings.production"
python manage.py makemigrations
python manage.py migrate
# Start Gunicorn processes
echo "Starting Gunicorn."
exec gunicorn edatool.wsgi:application \
    --bind 0.0.0.0:8000 \
    --workers 3

If you only want to use Dockerfile, you can add ENTRYPOINT[] command. Example how to run .sh script:

FROM python:3.9.4
RUN apt-get update
RUN apt-get install libpq-dev --assume-yes
RUN pip3 install psycopg2

COPY . /app
WORKDIR /app

RUN pip install -r requirements.txt
RUN pip3 install debugpy

ENTRYPOINT ["/app/docker-entrypoint.sh"]

CMD ["python", "manage.py", "runserver", "0.0.0.0:8000"]