[FLINK-32111] Replacing cluster in failed state with a new one failed

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: kubernetes-operator-1.5.0, kubernetes-operator-1.6.0
• Fix Version/s: kubernetes-operator-1.6.0
• Component/s: Kubernetes Operator
• Assignee: Tamir Sagi
• Reporter: Tamir Sagi
• Created: 16/May/23 15:07
• Updated: 15/Jun/23 11:40
• Resolved: 19/May/23 15:43

Description

I deployed a problematic cluster(HA enabled with 3 JMs) to check cluster updates process. The cluster was in restart loops. Then I provided a newer CRD (Updated several configurations) and expected the cluster to get re-deployed. however the following exception happened

 

Caused by: java.lang.NullPointerException
        at org.apache.flink.kubernetes.operator.service.CheckpointHistoryWrapper.getInProgressCheckpoint(CheckpointHistoryWrapper.java:60) 
        at org.apache.flink.kubernetes.operator.service.AbstractFlinkService.getCheckpointInfo(AbstractFlinkService.java:564) 
        at org.apache.flink.kubernetes.operator.service.AbstractFlinkService.getLastCheckpoint(AbstractFlinkService.java:520) 
        at org.apache.flink.kubernetes.operator.observer.SavepointObserver.observeLatestSavepoint(SavepointObserver.java:209) 
        at org.apache.flink.kubernetes.operator.observer.SavepointObserver.observeSavepointStatus(SavepointObserver.java:73) 
        at org.apache.flink.kubernetes.operator.observer.deployment.ApplicationObserver.observeFlinkCluster(ApplicationObserver.java:61) 
        at org.apache.flink.kubernetes.operator.observer.deployment.AbstractFlinkDeploymentObserver.observeInternal(AbstractFlinkDeploymentObserver.java:73) 
        at org.apache.flink.kubernetes.operator.observer.AbstractFlinkResourceObserver.observe(AbstractFlinkResourceObserver.java:53) 
        at org.apache.flink.kubernetes.operator.controller.FlinkDeploymentController.reconcile(FlinkDeploymentController.java:134) 

 

upgradeMode was first `last-state` and then I changed it to `stateless` but it still did not deploy the new cluster.

Issue Links

• is duplicated by: FLINK-32340 NPE in K8s operator which brakes current and subsequent deployments