Conversation
This is completely untested but will be the basis of experimenting with Claude in a container to follow all of the directions and iterate on the document until it works end to end.
| .. note:: | ||
|
|
||
| **Container environments:** If running in a Docker container without systemd, use | ||
| ``sudo service mysql start`` instead of systemctl commands. |
There was a problem hiding this comment.
I suspect this is Claude being confused about something it tried to run, we should probably just pick service and try not to worry about the layers underneath
| redis-cli ping | ||
| # Should return: PONG | ||
|
|
||
| 5. Install and Configure Memcached |
There was a problem hiding this comment.
Not sure that we should worry about both memcached and redis, should probably just focus on a redis install at this point?
| pip list | grep -i celery | ||
| # Should show Celery | ||
|
|
||
| 11. Install Node.js Dependencies and Build Assets |
There was a problem hiding this comment.
This is repeated at step 17, I suspect it got confused and didn't know to copy the files so did it twice?
| # Create superuser with --noinput (non-interactive) | ||
| ./manage.py lms createsuperuser --username admin --email admin@example.com --noinput --settings=production | ||
|
|
||
| # Set the password using Python |
There was a problem hiding this comment.
This is silly, I think we can just set the DJANGO_SUPERUSER_EMAIL / DJANGO_SUPERUSER_PASSWORD / DJANGO_SUPERUSER_USERNAME env vars
bmtcril
left a comment
bmtcril
left a comment
There was a problem hiding this comment.
Still some work to do, but want to get other eyes on it
|
Closing / reopning to kick RtD |
|
@bmtcril just FYI it seems like RTD is taking about 10 minutes to report build status back to GH. Clicking into the build you can see status & view the built docs. |
|
@sarina yeah, yesterday RtD was dying in the middle of building and failed to report status back at all |
2 participants
This is a first pass doc authored iteratively with Claude Code in a fresh Ubuntu 24.04 Docker container. The different commits show the progress as we worked through various install pitfalls and eventually got to a working build with the essential MFEs all connecting. As near as I can tell everything was running, but it has not been thoroughly tested.
An important note is that I did no installation or configuration work. I set up Claude Code, told it to write the install doc, then had it perform all of the steps, fixing the doc as it went. A record of the final session is here.
There are large chunks of leftover documentation about the process and testing at the bottom, which I think are actually really useful for hunting down common issues and I'd argue that they should stay in one form or another.
Known issues: