root
/
kafka
mirror of https://github.com/apache/kafka.git
1
0
Fork 0
Code Issues Actions 2 Packages Projects Releases Wiki Activity

KAFKA-18771: fix Flaky test KRaftClusterTest .testDescribeQuorumRequestToControllers (#18859) 

The case testDescribeQuorumRequestToControllers shutdowns raft client but not the controller. This makes client has chance to send a request to the controller and get NOT_LEADER_OR_FOLLOWER error. However, if the raft client finishes shutdown before handling the request, the request will not be handled. Shutdown the controller before doing KafkaFuture#get for the client request, so we can make sure the request is handled by another controller eventually.

Signed-off-by: PoAn Yang <payang@apache.org>

Reviewers: Luke Chen <showuon@gmail.com>
This commit is contained in:
PoAn Yang 2025-02-12 16:16:43 +08:00 committed by GitHub
parent 400363b7e2
commit 63fc9b3cb8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
core/src/test/scala/integration/kafka/server/KRaftClusterTest.scala
View File

@ -962,11 +962,16 @@ class KRaftClusterTest {
s"Leader ID ${quorumInfo.leaderId} was not a controller ID.") s"Leader ID ${quorumInfo.leaderId} was not a controller ID.")
// Try to bring down the raft client in the active controller node to force the leader election. // Try to bring down the raft client in the active controller node to force the leader election.
// Stop raft client but not the controller, because we would like to get NOT_LEADER_OR_FOLLOWER error first.
// If the controller is shutdown, the client can't send request to the original leader.
cluster.controllers().get(quorumInfo.leaderId).sharedServer.raftManager.client.shutdown(1000) cluster.controllers().get(quorumInfo.leaderId).sharedServer.raftManager.client.shutdown(1000)
// Send another describe metadata quorum request, it'll get NOT_LEADER_OR_FOLLOWER error first and then re-retrieve the metadata update // Send another describe metadata quorum request, it'll get NOT_LEADER_OR_FOLLOWER error first and then re-retrieve the metadata update
// and send to the correct active controller. // and send to the correct active controller.
val quorumInfo2 = admin.describeMetadataQuorum(new DescribeMetadataQuorumOptions) val quorumInfo2Future = admin.describeMetadataQuorum(new DescribeMetadataQuorumOptions).quorumInfo
.quorumInfo().get() // If raft client finishes shutdown before returning NOT_LEADER_OR_FOLLOWER error, the request will not be handled.
// This makes test fail. Shutdown the controller to make sure the request is handled by another controller.
cluster.controllers.get(quorumInfo.leaderId).shutdown()
val quorumInfo2 = quorumInfo2Future.get
// Make sure the leader has changed // Make sure the leader has changed
assertTrue(quorumInfo.leaderId() != quorumInfo2.leaderId()) assertTrue(quorumInfo.leaderId() != quorumInfo2.leaderId())