[
https://issues.apache.org/jira/browse/KAFKA-12462?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17300650#comment-17300650
]
A. Sophie Blee-Goldman commented on KAFKA-12462:
------------------------------------------------
Thanks Walker! This actually seems like a long-lurking bug that was just
surfaced by the removeStreamThread() feature, not caused by it. Before we could
remove threads this was only possible when shutting down the client, which we
don’t test as frequently as we now do removeStreamThread(). It’s also hard to
notice that a bug has caused thread(s) to die when the threads were supposed to
shut down anyways. But now we might only be removing one thread, and thanks to
the new exception handler we’ll shut down the whole application upon hitting
this so the thread won’t just quietly die.
We should consider backporting the fix to 2.7, even though the bug isn't going
to be as frequent or as bad in earlier versions. I wouldn't cut a new RC for
2.6.2 over this, but we might as well backport to get the fix in 2.7.1 whenever
that comes out
> Threads in PENDING_SHUTDOWN entering a rebalance can cause an illegal state
> exception
> --------------------------------------------------------------------------------------
>
> Key: KAFKA-12462
> URL: https://issues.apache.org/jira/browse/KAFKA-12462
> Project: Kafka
> Issue Type: Bug
> Reporter: Walker Carlson
> Priority: Blocker
> Fix For: 2.8.0
>
>
> A thread was removed, sending it to the PENDING_SHUTDOWN state, but went
> through a rebalance before completing the shutdown.
> {code:java}
> // [2021-03-07 04:33:39,385] DEBUG [i-07430efc31ad166b7-StreamThread-6]
> stream-thread [i-07430efc31ad166b7-StreamThread-6] Ignoring request to
> transit from PENDING_SHUTDOWN to PARTITIONS_REVOKED: only DEAD state is a
> valid next state (org.apache.kafka.streams.processor.internals.StreamThread)
> {code}
> Inside StreamsRebalanceListener#onPartitionsRevoked, we have
> {code:java}
> //
> if (streamThread.setState(State.PARTITIONS_REVOKED) != null &&
> !partitions.isEmpty())
> taskManager.handleRevocation(partitions);
> {code}
> Since PENDING_SHUTDOWN → PARTITIONS_REVOKED is a disallowed transition, we
> never invoke TaskManager#handleRevocation. Currently handleRevocation is
> responsible for preparing any active tasks for close, including committing
> offsets and writing the checkpoint as well as suspending the task. We can’t
> close the task in handleRevocation since we still support EAGER rebalancing,
> which invokes handleRevocation at the beginning of a rebalance on all tasks.
> The tasks that are actually revoked will be closed during
> TaskManager#handleAssignment . The IllegalStateException is specifically
> because we don’t suspend the task before attempting to close it, and the
> direct transition from RUNNING → CLOSED is forbidden.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
