[FLINK-21735] Harden JobMaster#updateTaskExecutionState()

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.13.0
• Fix Version/s: 1.11.4, 1.13.0, 1.12.3
• Component/s: Runtime / Coordination
• Labels: pull-request-available
• Assignee: Nicholas Jiang
• Reporter: Zhu Zhu
• Created: 11/Mar/21 12:39
• Updated: 15/Mar/21 12:59
• Resolved: 15/Mar/21 12:59

Description

Currently JobMaster#updateTaskExecutionState() is very important to trigger task scheduling and recovery. We do not expect any exception to happen directly in its invocation, so I'd propose to wrap it with a try-catch to fail the JobMaster once any exception is caught. This can better expose the unexpected bugs.

Issue Links

• is related to: FLINK-21707 Job is possible to hang when restarting a FINISHED task with POINTWISE BLOCKING consumers
• links to: GitHub Pull Request #15196