gevent cooperative multitasking appears to break Connexion v3 FlaskApp?

[chrisinmtown]

I am trying to upgrade an app from Connexion v2 to v3. The old app uses the FlaskApp and gevent-provided cooperative multi-tasking so the server can process multiple requests in parallel. The old approach of calling the gevent.patch_all() function at start seems to break connexion. From what I can see, requests to the Connexion server don't finish, they just hang.

To reproduce the behavior, find the Connexion helloworld example and add the following as the first line in hello.py:

import gevent.monkey; gevent.monkey.patch_all()

Install the prereqs and starting the server:

pip install -r requirements.txt
python hello.py

The startup takes a second longer than before, but no complaints appear. Then try to load the Swagger UI at http://127.0.0.1:8080/openapi/ui/

In my experience this /never finishes/ it just loads forever. Then hit Ctrl-C to interrupt the server. It responds Shutting down but hangs again, a second Ctrl-C is required to make it stop.

Versions: Python 3.12, connexion 3.1.0, flask 3.1.0 , gevent 24.11.1.

Is this enough information? Please advise, thanks in advance.

[chrisinmtown]

I hope someone will comment on the breakage. Maybe the connexion migration guide should mention this? I expect I'm not the only person trying to migrate a Connexion v2 app who chose to use gevent like this.

I tried using gevent bcos I would like my connexion v3 app to process more than 1 request concurrently in production where I use uvicorn. I understand now that uvicorn provides a basic process manager, and spawns processes according to the workers parameter supplied to the uvicorn.run() function. It appears that this is sufficient, and I don't need the gevent cooperative multitasking monkey patch.

[FelixSchwarz]

From my (very limited) experience, a lot of flask-centric implementations need a big overhaul when switching to Connexion 3.

That being said: Are you sure you really need gevent? (also with Connexion 2.x)
I am a fan of gevent because it allowed me to use "async with colorless functions" and use libraries which were not designed for asynchronous use.

However if you just want to process a few requests in parallel and you don't need all requests in the same process, there is the traditional approach of using gunicorn + multiple workers which is way less complex than using gevent.

My gut feeling is that instead of gevent, you could use a full async app with Connexion 3.