Ambiguous error message `metadata_unknown` with 422 status code when launching sessions

[olevski]

Users from the meteo swiss cluster reported seeing this when trying to launch a session on a public project, with a public github repository.

It seems that this can happen when we cannot reach github to get additional information about the repository. But the error message should be more descriptive, not just metadata_unknown. Even I had to dig through the code to see what this means. It seems that there is a an enum in the code that contains different erorr cases and conditions that in some cases gets piped straight into the error message without any disambiguation.

Seen on renku 2.17.1.

It may be that this is worse or more common when the repo is saved with the url pointing to the ui page only. I.e. like https://github.com/MeteoSwiss/opendata-nwp-demos rather than https://github.com/MeteoSwiss/opendata-nwp-demos.git. But I have not been able to reproduce this even when the repo url is not in the optimal form.

[olevski]

Here is how this error occured in production from the logs:

[ERROR] 22/MainThread renku_data_services./app/env/lib/python3.13/site-packages/renku_data_services/repositories/db.py (db.py:177) [06c2f08e-c106-cd63-75fe-4fa178a70075][ecbbb04f2c481b6b5fda913ebcffcf9e] - Error status 403 returned for repository metadata: https://api.github.com/repos/MeteoSwiss/opendata-nwp-demos

[ERROR] 22/MainThread sanic.error (error.py:201) [06c2f08e-c106-cd63-75fe-4fa178a70075][ecbbb04f2c481b6b5fda913ebcffcf9e] - Exception occurred while handling uri: 'https://renkulab.io/api/data/sessions'

Traceback (most recent call last):
  File "handle_request", line 99, in handle_request
    Extend = TypeVar("Extend", type)  # type: ignore
                       ^^^^^^^^^^^^^^	
  File "/app/env/lib/python3.13/site-packages/renku_data_services/base_api/auth.py", line 66, in decorated_function	
    response = await f(request, user1, user2, *args, **kwargs)	
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^	
  File "/app/env/lib/python3.13/site-packages/sanic_ext/extras/validation/decorator.py", line 79, in decorated_function	
    retval = await retval	
             ^^^^^^^^^^^^	
  File "/app/env/lib/python3.13/site-packages/renku_data_services/notebooks/blueprints.py", line 83, in _handler	
    session, created = await start_session(
                       ^^^^^^^^^^^^^^^^^^^^
    ...<19 lines>...
    )	
    ^	
  File "/app/env/lib/python3.13/site-packages/renku_data_services/notebooks/core_sessions.py", line 892, in start_session	
    raise errors.ValidationError(message=str(repo_data.error))
renku_data_services.errors.errors.ValidationError: ValidationError: metadata_unknown

[olevski]

Related to #1323. This is occurring because we are getting rate limited by Github. The other issue is about dealing with the rate limiting.

[leafty]

The repository check happened from #1250 as part of the build from private code repository. I feel it not necessary when the image is public (build from public repo) and I wanted to fix the check here -> #1283. But I didn't have the time to finish this. To be noted, the correct repo permission check when the image is in private harbor is far from trivial (need to be cross-checked against last successful build) which is why implementation was not completed.

[leafty]

Oh, and of course, the PR tries to fix the error message so that a proper 4xx (FORBIDDEN) is sent back, not a 422 validation error.