getsentry
Work around libraries/external component's logging message issues
Problem Statement
Using gunicorn's gevent worker requests sometimes time out during normal operation. In that case a warning level logging event is triggered.
Unfortunately the message is pre-formatted and includes an ever changing pid, which obviously breaks sentry's grouping and we end up with lots of un-grouped issues.
It would be interesting to see them grouped to analyze time of day or clients causing timeouts the most. Currently we need to manually merge them for this to be possible.
Solution Brainstorm
I have opend an issue with pull-request for gunicorn to fix the logging code.
A configuration/option to setup automatic merging rules would be nice to generally "fix" such issues with badly behaving libraries/external components.
Hey @fdemmer Thanks for reporting this and also creating a PR in Gunicorn. Fixing this in Gunicorn is probably the cleanest solution.
Let's see if they merge your PR.
As for the "grouping rules" idea: we are currently collecting information on how to improve the grouping in different languages/frameworks. So whenever you come across a situation where the grouping could be improved, please file a Issue here! Thanks!
As this was fixed in Gunicorn, this should be resolved now. Thanks everyone!