Document how to configure the connexion logger to see internal exceptions in logs

[xcelerit-team]

Description

Being used to Flask, we usually configured the default handler for flask's app.logger and that enabled us to see detailed information, including stack trace, for exceptions raised in a view function. This is very helpful after an internal server error occurred, to dig down to the cause in the logs.

With Connexion 3.0, the default exception handler uses connexion's own logger, and if that does not have a handler registered, all we can see is an internal server error response with no details. Especially in development this is very annoying, as the root cause cannot be found easily. This also applies to tests.

Expected behaviour

Enable logging to stderr for at least exception-level, or document clearly how to add a handler to connexion's own logger for this purpose.

Actual behaviour

No error details visible.

Steps to reproduce

Write any code in a view handler that raises an exception and run a test. Details of the actual error cannot be found by default.

[RobbeSneyders]

Hi @lotzej

The default exception handlers registered in Connexion should log any errors by default. I cannot reproduce the lack of a stack trace, both when running an application, or when running a test.

Connexion follows logging best practices, so you should just be able to get access to the connexion logger using:

import logging

connexion_logger = logging.getLogger("connexion")

Could you confirm if that addresses your use case? Then we can add this to the documentation.

[xcelerit-team]

Yes, accessing the logger like this and adding a handler to it that goes to standard output fixes the issue. Without that, I don't see a stack trace.

Would be good to document that - people used to flask would typically already have code to setup the flask app logger, and that worked previously. With Connexion 3.0, the exception logs go elsewhere, so people might miss that.

[xcelerit-team]

I might be doing something wrong, but if I don't touch the logging configuration in any way and just run tests, I don't see any exception traces logged.

Logging best practices says that warning level and above should go to stderr by default for a library.

[RobbeSneyders]

I see we do set the logging level in our tests.

I'm not sure why it's needed though. Connexion doesn't set any logging configuration, so I would expect the default level to be warning as you mention.

[xcelerit-team]

You're right actually - without logging configuration it shows the warnings/exceptions automatically. We had reconfigured the logging setting for flask in the previous version, and that removed them. It was a major headache, as that's easy to miss when upgrading to connexion 3.0. Previously the errors were showing (by flask) - now they were not.