docker entrypoint behaviour with django

Question

I'm trying to make my first django container with uwsgi. It works as follows:

FROM python:3.5

RUN apt-get update && \
    apt-get install -y && \
    pip3 install uwsgi

COPY ./projects.thux.it/requirements.txt /opt/app/requirements.txt
RUN pip3 install -r /opt/app/requirements.txt

COPY ./projects.thux.it /opt/app
COPY ./uwsgi.ini  /opt/app
COPY ./entrypoint /usr/local/bin/entrypoint

ENV PYTHONPATH=/opt/app:/opt/app/apps

WORKDIR /opt/app
ENTRYPOINT ["entrypoint"]
EXPOSE 8000

#CMD ["--ini", "/opt/app/uwsgi.ini"]

entrypoint here is a script that detects whether to call uwsgi (in case there are no args) or python manage in all other cases. I'd like to use this container both as an executable (dj migrate, dj shell, ... - dj here is python manage.py the handler for django interaction) and as a long-term container (uwsgi --ini uwsgi.ini). I use docker-compose as follows:

web:
    image: thux-projects:3.5
    build: .
    ports:
      - "8001:8000"
    volumes:
      - ./projects.thux.it/web/settings:/opt/app/web/settings
      - ./manage.py:/opt/app/manage.py
      - ./uwsgi.ini:/opt/app/uwsgi.ini
      - ./logs:/var/log/django

And I manage in fact to serve the project correctly but to interact with django to "check" I need to issue:

docker-compose exec web entrypoint check

while reading the docs I would have imagined I just needed the arguments (without entrypoint)

Command line arguments to docker run will be appended after all elements in an exec form ENTRYPOINT, and will override all elements specified using CMD. This allows arguments to be passed to the entry point, i.e., docker run -d will pass the -d argument to the entry point.

The working situation with "repeated" entrypoint:

$ docker-compose exec web entrypoint check
System check identified no issues (0 silenced).

The failing one if I avoid 'entrypoint':

$ docker-compose exec web check
OCI runtime exec failed: exec failed: container_linux.go:346: starting container process caused "exec: \"check\": executable file not found in $PATH": unknown

Answer 1

docker exec never uses a container's entrypoint; it just directly runs the command you give it.

When you docker run a container, the entrypoint and command you give to start it are combined to produce a single command line, and that command becomes the main container process. On the other hand, when you docker exec a command in a running container, it's interpreted literally; there aren't two parts of the command line to assemble, and the container's entrypoint isn't considered at all.

For the use case you describe, you don't need an entrypoint script to process the command in an unusual way. You can create a symlink to the manage.py script to give a shorter alias to run it, but make the default command be the uwsgi runner.

RUN chmod +x manage.py
RUN ln -s /opt/app/manage.py /usr/local/bin/dj
CMD ["uwsgi", "--ini", "/opt/app/uwsgi.ini"]

# Runs uwsgi:
docker run -p 8000:8000 myimage

# Manually trigger database migrations:
docker run --rm myimage dj migrate