Failing check runs causes erroneous scale down #3584
Labels
bug
Something isn't working
community
Community contribution
needs triage
Requires review from the maintainers
Checks
Controller Version
0.27.6
Helm Chart Version
0.23.7
CertManager Version
1,12,2
Deployment Method
Helm
cert-manager installation
Not a problem with cert-manager.
Checks
Resource Definitions
To Reproduce
Not applicable, see #2118
Describe the bug
This bug is a continuation of issue #2118 more specifically the fixes implemented in PR #2520.
The fix tries to exclude check runs from causing scale down, but as can be seen on line 220 of the fix, it only does so if the check is successful:
actions-runner-controller/controllers/actions.summerwind.net/horizontal_runner_autoscaler_webhook.go
Line 220 in 8947327
By the nature of check runs, they are intended to fail if there's a problem. So I question why this filter only applies for the "success" conclusion. As it is now, it causes erroneous scale down whenever a check fails (yes. like in the original issue, we're using EnricoMi/publish-unit-test-result-action)
One observation I have is that these events have the
labels
set to[]
so perhaps that would be a better way to detect them?Describe the expected behavior
I expect check run events to not cause scale down regardless of the conclusion.
Whole Controller Logs
Not applicable, see #2118
Whole Runner Pod Logs
Not applicable, see #2118
Additional Context
No response
The text was updated successfully, but these errors were encountered: