[FLINK-25486] Perjob can not recover from checkpoint when zookeeper leader changes

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.13.5, 1.14.2, 1.15.0
• Fix Version/s: 1.13.6, 1.14.4, 1.15.0
• Component/s: Runtime / Coordination
• Labels: pull-request-available
• Assignee: Liu
• Reporter: Liu
• Created: 30/Dec/21 11:20
• Updated: 30/Jan/22 16:23
• Resolved: 29/Jan/22 15:45

Description

When the config high-availability.zookeeper.client.tolerate-suspended-connections is default false, the appMaster will failover once zk leader changes. In this case, the old appMaster will clean up all the zk info and the new appMaster will not recover from the latest checkpoint.

The process is as following:

1. Start a perJob application.
2. kill zk's leade node which cause the perJob to suspend.
3. In MiniDispatcher's function jobReachedTerminalState, shutDownFuture is set to UNKNOWN .
4. The future is transferred to ClusterEntrypoint, the method is called with cleanupHaData true.
5. Clean up zk data and exit.
6. The new appMaster will not find any checkpoints to start and the state is lost.

Since the job can recover automatically when the zk leader changes, it is reasonable to keep zk info for the coming recovery.

 

Issue Links

• links to: GitHub Pull Request #18296