-
Bug
-
Resolution: Done
-
Highest
-
Dublin Release
-
None
-
SB-01
The SDC-BE pod liveness probe fails when the pod finds it loses connectivity to Cassandra. As the consequence the pod is restarted by k8s and won't be available until it's fully up. This happens often when there is small packet loss in network.
The issue here is the pod liveness should not depend on external Cassandra. The connectivity issue could recover quickly by itself, we can't afford the longer period of pod restart just because of transient connectivity issue (in that case restarting the pod does not help)