500 errors when accessing knowledge repo through gunicorn deploy

[rcquan]

I am able to successfully use the runserver command to start a development server and view the knowledge repo app through my web browser.

However, when I use the deploy command and kick off gunicorn workers successfully, I am unable to load any pages of the knowledge repo in my web browser and instead receive a 500 response.

Unfortunately, I am unable to provide a useful stack trace since the application logger doesn't seem to print to stdout/stderr when using gunicorn.

[matthewwardrop]

Hi @rcquan! Thanks for letting us know about this. Can you try the latest code (we fixed quite a few bugs on Friday)?

If that doesn't work, can you run:
knowledge_repo --repo <repo_path> --debug deploy

That should enable you to see tracebacks.

[matthewwardrop]

An update on this: depending on whether you were using multiple repositories via the MetaKnowledgeRepository multiplexer, there were some issues regarding gunicorn deployment that have been fixed in #33 . Can you confirm that you are still having issues @rcquan ?

[rcquan]

The latest changes seems to have fixed the issue. I just had to delete the .resources directory and re-initialize the current project with knowledge_repo init in order to get the new changes to apply to my project. Thanks @matthewwardrop 👍

[matthewwardrop]

Ah... Excellent! Thanks for keeping this up to date @rcquan :). It should not have been necessary to re-initialize like this... but I'm glad it worked at the end.

[matthewwardrop]

@rcquan Oh... wait! This is exactly what should have happened regarding the embedded tooling. The idea is that the embedded tools are frozen at a particular revision, so that all users sharing the repository are using a tested sanctioned version. It also allows you to update the version of the tools used by everyone using your repository without them having to pip install knowledge-repo again. You manually update the submodule index in order to update the embedded tools. If you are using it just for yourself, or for a small group of people, you can skip embedding the tools (use --tooling-skip during the init phase), or use --dev to bypass chaining the knowledge_repo command to the embedded version of the tools.

Hope this makes sense.

[rcquan]

@matthewwardrop Ah, that makes sense! I hadn't realized that the technical details section of the doc also contained information on how to update the knowledge-repo version for the repository. I imagine this will be particularly important point to highlight as y'all bump new versions in the beta ;)