Revert "loadbalancer: Keep non-serving terminating backends"#41199
Revert "loadbalancer: Keep non-serving terminating backends"#41199jrife wants to merge 1 commit intocilium:mainfrom
Conversation
This reverts commit 6f41c98. Related: cilium#41194 Signed-off-by: Jordan Rife <jrife@google.com>
|
/test |
|
I still don't see any connection whatsoever with the commit and the failing test (the failing test doesn't involve terminating endpoints). I'll rerun the ginkgo test a few more times here with the revert to see if it can be triggered without this commit. EDIT: The one potential problem could be that we're seeing endpoints without conditions leading them to be quarantined and on agent restart we're picking up this quarantined state that is retained until backend is health checked (which only happens when you have a health checker, which you don't in these tests). We can remove the restoration of the quarantined state as now we default to unhealthy if a health checker is registered. To test if this is the case I opened #41234 to just ignore endpoints that have no conditions and thus won't get quarantined and will do few runs of ginkgo tests there. |
|
#41234 solved the flakyness. Closing this one. |
Pull request was closed
This reverts commit 6f41c98.
Related: #41194
ci-ginkgo has been failing consistently and it seems to have started with this commit.