[FLINK-10129] Flink job IDs are not getting deleted automatically from zookeeper metadata after canceling flink job in flink HA cluster

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Duplicate
• Assignee: Unassigned
• Reporter: Keshav Lodhi
• Created: 13/Aug/18 06:48
• Updated: 14/Sep/18 09:25
• Resolved: 14/Sep/18 09:25

Description

Hi Team,

Here is, what i am looking for:

• We have  flink HA dockerized cluster with (3 zookeepers, 2 job-managers, 3 task-managers) 
• So whenever we are cancelling the flink job, it is getting cancelled but it is not deleting the cancelled job ID from the zookeeper metadata (Inside flink/jobgraph folder in zookeeper) automatically. 
• So whenever any one of the job-manager goes down/restarted , it doesn't come up and throws exception like  "Could not find this job id xxxxxxxxxx".
• The current work around is to remove the canceled job ID from the zookeeper metadata manually. (But this is not the recommended solution).     

 

Please advise.

Issue Links

• duplicates: FLINK-10255 Standby Dispatcher locks submitted JobGraphs