-
Bug
-
Resolution: Unresolved
-
Critical
-
None
-
Jenkins ver. 2.180
Kubernetes-plugin ver 1.16.0
Kubernetes ver 1.12.7
Zombie pods are created when resources are not sufficient for secluding pods and idleMinutes is set to a non-zero value.
This causes multiple pods to be launched from one pod template where one of the pods is in a zombie state and one is active. As the zombie pods accumulate, there are no resources left over to schedule actual pods.
WAR: set up a cron job to delete pods in error state or set idleMinutes to 0.
Is there a simple way to reproduce this condition from scratch?