[FLINK-35522] The source task may get stuck after a failover occurs in batch jobs

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.17.2, 1.19.0, 1.18.1, 1.20.0
• Fix Version/s: 1.18.2, 1.20.0, 1.19.2
• Component/s: Runtime / Coordination
• Labels: pull-request-available
• Assignee: xingbe
• Reporter: xingbe
• Created: 05/Jun/24 01:22
• Updated: 07/Jun/24 02:00
• Resolved: 07/Jun/24 02:00

Description

If the source task does not get assigned a split because the SplitEnumerator has no more splits, and a failover occurs during the closing process, the SourceCoordinatorContext will not resend the NoMoreSplit event to the newly started source task, causing the source vertex to remain stuck indefinitely. 
This case may only occur in batch jobs where speculative execution has been enabled.

Issue Links

• links to: GitHub Pull Request #24887
• links to: GitHub Pull Request #24898
• links to: GitHub Pull Request #24899