rondagostino commented on a change in pull request #10003:
URL: https://github.com/apache/kafka/pull/10003#discussion_r567468635
##########
File path: core/src/main/scala/kafka/server/ReplicaManager.scala
##########
@@ -257,6 +265,25 @@ class ReplicaManager(val config: KafkaConfig,
}
}
+ // Changes are initially deferrable when using a Raft-based metadata quorum,
and may flip-flop thereafter;
+ // changes are never deferrable when using ZooKeeper. When true, this
indicates that we should transition online
+ // partitions to the deferred state if we see metadata with a different
leader epoch.
+ @volatile private var changesDeferrable: Boolean = usingRaftMetadataQuorum
+ stateChangeLogger.info(s"Metadata changes deferrable=$changesDeferrable")
+
+ private def confirmChangesNotDeferrableWithZooKeeper(): Unit = {
+ if (changesDeferrable) {
+ throw new IllegalStateException("Partition metadata changes should never
be deferrable when using ZooKeeper")
+ }
+ }
+
+ def deferrableMetadataChanges(): Unit = {
+ replicaStateChangeLock synchronized {
Review comment:
It isn't entirely clear since there are places where this lock is
acquired that we won't implement for the initial version of a Raft-based
metadata quorum (e.g. handling log dir failure). I think it's a good point
that we should check here, though, because we must never make changes
deferrable if we are using ZooKeeper. So I added the check.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
