mumrah commented on code in PR #13407:
URL: https://github.com/apache/kafka/pull/13407#discussion_r1173116859
##########
metadata/src/main/java/org/apache/kafka/controller/QuorumController.java:
##########
@@ -1123,29 +1165,104 @@ private void claim(int epoch) {
// Prepend the activate event. It is important that this event go
at the beginning
// of the queue rather than the end (hence prepend rather than
append). It's also
// important not to use prepend for anything else, to preserve the
ordering here.
- queue.prepend(new ControllerWriteEvent<>("completeActivation[" +
epoch + "]",
- new CompleteActivationEvent()));
+ ControllerWriteEvent<Void> activationEvent = new
ControllerWriteEvent<>("completeActivation[" + epoch + "]",
+ new CompleteActivationEvent(),
+ EnumSet.of(DOES_NOT_UPDATE_QUEUE_TIME,
RUNS_IN_PREMIGRATION));
+ activationEvent.future.whenComplete((__, t) -> {
+ if (t != null) {
+ fatalFaultHandler.handleFault("exception while activating
controller", t);
Review Comment:
In `handleEventException`, we renounce the controller leadership when we see
a non-ApiException from a ControllerWriteEvent. As far as I can tell, this does
not hit the fault handler.
`KafkaEventQueue#run` traps exceptions and calls `Event#handleException`. In
`ControllerEvent#handleException`, we call
`QuorumController#handleEventException` which just does logging and
resignation. `handleEventException` wraps the exception with
UnknownServerException, but I don't see where we do anything with that in
ControllerEvent. I think this is by design since we don't want the controller
just crashing for any error, but only for specific cases where we call the
fault handler.
--
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
