Bugfix - Production Django Container Restart Loop

[studebacon]

The production django container would get stuck in a restart loop due do a lack of permission on the /app/ghostwriter/media folder.

Running mkdir /app/ghostwriter/media prior to chown -R django /app in the Dockerfile corrects this issue.

[Kaicastledine]

Had the same issue. Logs found:

mkdir: can't create directory '/app/ghostwriter/media/evidence': Permission denied
mkdir: can't create directory '/app/ghostwriter/media/user_avatars': Permission denied
mkdir: can't create directory '/app/ghostwriter/media/templates': Permission denied
cp: can't create '/app/ghostwriter/media/templates': Permission denied
PostgreSQL is available
mkdir: can't create directory '/app/ghostwriter/media/evidence': Permission denied
mkdir: can't create directory '/app/ghostwriter/media/user_avatars': Permission denied
mkdir: can't create directory '/app/ghostwriter/media/templates': Permission denied

Symptons:

bca6682deff8   ghostwriter_production_nginx      "/docker-entrypoint.…"   14 seconds ago   Restarting (1) 3 seconds ago              ghostwriter_nginx_1
e58913957285   ghostwriter_production_django     "/entrypoint /start"     15 seconds ago   Restarting (1) 1 second ago               ghostwriter_django_1

Updating the dockerfile for django as instructed above fixed the issue

[chrismaddalena]

@studebacon @Kaicastledine Someone else mentioned an issue (off GitHub) that could be related to this. I looked into with them and we determined the container had built incorrectly. I have not encountered it myself, but this seems to be a rare issue caused by the build going a bit sideways.

I'm not opposed to working around a rare permissions snafu with a chown, but I am curious if this started occurring out of the blue and when.

[Kaicastledine]

@studebacon @Kaicastledine Someone else mentioned an issue (off GitHub) that could be related to this. I looked into with them and we determined the container had built incorrectly. I have not encountered it myself, but this seems to be a rare issue caused by the build going a bit sideways.

I'm not opposed to working around a rare permissions snafu with a chown, but I am curious if this started occurring out of the blue and when.

So this was a new issue with the last release, if that's any help.

[chrismaddalena]

Yep! That does help. With the latest release, we bumped up versions of a few things, including the Alpine Linux image. That might explain the appearance of a rare filesystem bug like this. This PR is likely the best way for the project to address it.

[arledesma]

This is a race condition with the nginx container. The nginx container is defaulted to USER root, mounting the volume before django, and subsequently the docker-compose shorthand syntax for volumes is not set for read_only or volume: nocopy.

[chrismaddalena]

That's interesting, @arledesma. It makes sense. A race condition would explain why it's not easily reproduced and wasn't reported by others. I included this patch in the latest release, v2.2, but I'm curious to see where your WIP #163 goes.