junrao commented on a change in pull request #11255:
URL: https://github.com/apache/kafka/pull/11255#discussion_r698652526
##########
File path: core/src/main/scala/kafka/controller/KafkaController.scala
##########
@@ -628,9 +628,9 @@ class KafkaController(val config: KafkaConfig,
topicDeletionManager.failReplicaDeletion(newOfflineReplicasForDeletion)
}
- // If replica failure did not require leader re-election, inform brokers
of the offline brokers
+ // If no partitions are affected, inform brokers of the offline brokers
// Note that during leader re-election, brokers update their metadata
- if (partitionsWithOfflineLeader.isEmpty) {
+ if (newOfflineReplicas.isEmpty ||
(newOfflineReplicasNotForDeletion.isEmpty &&
partitionsWithOfflineLeader.isEmpty)) {
Review comment:
On second thought, I think this change still doesn't cover all the
cases. The main issue is that in
partitionStateMachine.triggerOnlinePartitionStateChange(), we only send
updateMetadataRequest for partitions with successful leader change. If no
partition has successful leader change, no updateMetadataRequest will be sent.
The reason could be that newOfflineReplicas is empty, but it could also be that
the leader can't be elected for other reasons (e.g. no replica is live).
In PartitionStateMachine.doElectLeaderForPartitions(), we have 3 categories
of partitions, finishedUpdates, failedElections and updatesToRetry. Perhaps we
could return finishedUpdates all the way to
partitionStateMachine.triggerOnlinePartitionStateChange() and force a send of
updateMetadataRequest if finishedUpdates is empty.
##########
File path: core/src/main/scala/kafka/controller/KafkaController.scala
##########
@@ -628,9 +628,9 @@ class KafkaController(val config: KafkaConfig,
topicDeletionManager.failReplicaDeletion(newOfflineReplicasForDeletion)
}
- // If replica failure did not require leader re-election, inform brokers
of the offline brokers
+ // If no partitions are affected, inform brokers of the offline brokers
// Note that during leader re-election, brokers update their metadata
- if (partitionsWithOfflineLeader.isEmpty) {
+ if (newOfflineReplicas.isEmpty ||
(newOfflineReplicasNotForDeletion.isEmpty &&
partitionsWithOfflineLeader.isEmpty)) {
Review comment:
I think we only need to check the return result of
`partitionStateMachine.triggerOnlinePartitionStateChange()`.
`partitionStateMachine.handleStateChanges(partitionsWithOfflineLeader.toSeq,
OfflinePartition)` doesn't generate any updateMetadataRequest.
--
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
