ableegoldman commented on code in PR #19400:
URL: https://github.com/apache/kafka/pull/19400#discussion_r2059386588
##########
streams/src/main/java/org/apache/kafka/streams/processor/internals/StreamThread.java:
##########
@@ -875,7 +878,7 @@ public void run() {
streamsUncaughtExceptionHandler.accept(e, false);
// Note: the above call currently rethrows the exception, so
nothing below this line will be executed
} finally {
- completeShutdown(cleanRun);
+ completeShutdown(cleanRun, true);
Review Comment:
ah sorry i totally missed this question on my first pass 🤦♀️ No, we
actually do want to stay in the group and retain the existing default behavior
for Streams
if you're curious why the default behavior for Streams is the opposite of
the plain consumer client, it's because Streams is stateful which means
rebalancing and moving partitions can be a much heavier operation and involve
restoring many records from the changelog topics when partitions are shuffled
around. By skipping the leave group, we can often avoid an unnecessary extra
rebalance that would result in partitions being moved, since the assignment of
Streams is heavily dependent on the current group members. If an error gets
thrown and the StreamThread dies and gets restarted within the session timeout,
we can avoid any rebalances where the dead thread is missing, so the rebalance
when the new thread restarts should result in the same assignment of partitions
to threads as we had before the thread died.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
