-
Bug
-
Resolution: Done
-
Normal
-
None
-
None
-
None
-
None
-
False
-
-
False
TestAutoscaleSustaining/aggregation-weightedExponential sometimes fails.
Two of the suspicious culprit are the following failure of PodSandBox creation.
serving-tests 16m Warning FailedCreatePodSandBox pod/autoscale-sustaining-aggregatio-yoztlumy-00001-deployment-w4s9r Failed to create pod sandbox: rpc error: code = Unknown desc = failed to create pod network sandbox k8s_autoscale-sustaining-aggregatio-yoztlumy-00001-deployment-w4s9r_serving-tests_97c243ff-31e6-4d34-ab74-a7be8113f985_0(ee5d5ba3884262b6a86d63b6f13411c7fb2da76f8a5b124cf212080b69aef64f): Multus: [serving-tests/autoscale-sustaining-aggregatio-yoztlumy-00001-deployment-w4s9r]: error setting the networks status: SetNetworkStatus: failed to update the pod autoscale-sustaining-aggregatio-yoztlumy-00001-deployment-w4s9r in out of cluster comm: SetNetworkStatus: failed to update the pod autoscale-sustaining-aggregatio-yoztlumy-00001-deployment-w4s9r in out of cluster comm: status update failed for pod /: resource name may not be empty
and
serving-tests 15m Warning FailedCreatePodSandBox pod/autoscale-sustaining-aggregatio-yoztlumy-00001-deployment-gd6n7 Failed to create pod sandbox: rpc error: code = Unknown desc = error reading container (probably exited) json message: EOF
The first error was fixed in OCP 4.8.2 by bz1957856
(please also refer to comment#8 in the bz. The condition is match to the error.)
"resource name may not be empty" error message is pretty common because it is client-go message. This bz only mentioned error message that is:
- Multus generates the this error message
- "pod /" is contained (this means pod/podnamespace is empty, pretty weird)
The second error was fixed in OCP 4.7 by bz1915085.
So, 4.6 CI could hit both error.
I open this Jira to see if we will hit the flake by any other reason or not.