Migrate to gunicorn + gevent

[robhudson]

No description provided.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 17 lines in your changes are missing coverage. Please review.

Project coverage is 87.24%. Comparing base (f945e17) to head (f01aaee).

Files	Patch %	Lines
basket/wsgi_config.py	0.00%	17 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1350      +/-   ##
==========================================
- Coverage   87.82%   87.24%   -0.58%     
==========================================
  Files          42       43       +1     
  Lines        2571     2588      +17     
==========================================
  Hits         2258     2258              
- Misses        313      330      +17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[robhudson]

I'm seeing some Python 3.12 + gevent issues:

RuntimeError: can't create new thread at interpreter shutdown

Python 3.12 no longer allows creating a thread on interpreter shutdown:
python/cpython#104826

[stevejalim]

r+ but why are we rolling this away from uwsgi? I half remember some niggles, which is presumably where gunicorn+gevent (rather than gunicorn+meinheld) is overall more stable, yeah?

Makes me wonder if we should also move Nucleus to match - I think we should

[robhudson]

Basket has been a good project to be the "lead", so to speak, on some library choices in the past. Since it's smaller and less complex. pmac and I discussed migrating basket to match bedrock so they are more in sync.

In some ways, this has already been beneficial since we've just discovered an issue with gevent workers under Python 3.12 due to the above mentioned change. So it's something to solve for bedrock (and basket, assuming we agree this is a good change).

[robhudson]

Historical note: We switched to granian on bedrock so closing this for now until the dust settles.