[FLINK-15301] Flink Kinesis AsyncRecordEmitter needs to handle unchecked exception gracefully

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.10.0
• Component/s: Connectors / Kinesis
• Labels: pull-request-available
• Assignee: Thomas Weise
• Reporter: Ying Xu
• Created: 17/Dec/19 17:41
• Updated: 31/Dec/19 01:24
• Resolved: 20/Dec/19 16:19

Description

Currently, any runTime exception encountered inside the `AsyncRecordEmitter.emitRecordAndUpdateState()` function could cause the thread to exit silently. Flink job would continue to run, but the stopped record emitter would subsequently cause Kinesis data consumption to stall. 

 

The AsyncRecordEmitter need to catch unchecked exception, log errors, and perhaps trigger job restart subsequently. 

Issue Links

• links to: GitHub Pull Request #10611
• links to: GitHub Pull Request #10725