[KAFKA-12474] Worker can die if unable to write new session key

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.4.2, 2.5.2, 2.8.0, 2.7.1, 2.6.2, 3.0.0
• Fix Version/s: 2.5.2, 2.8.0, 2.7.1, 2.6.2, 3.0.0
• Component/s: connect
• Assignee: Chris Egerton
• Reporter: Chris Egerton
• Reviewer: Randall Hauch
• Created: 16/Mar/21 03:27
• Updated: 01/Apr/21 23:22
• Resolved: 01/Apr/21 19:56

Description

If a distributed worker is unable to write (and then read back) a new session key to the config topic, an uncaught exception will be thrown from its herder's tick thread, killing the worker.

See https://github.com/apache/kafka/blob/8da65936d7fc53d24c665c0d01893d25a430933b/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/distributed/DistributedHerder.java#L366-L369

One way we can handle this case by forcing a read to the end of the config topic whenever an attempt to write a new session key to the config topic fails.

Issue Links

• links to: GitHub Pull Request #10396