You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "dajac (via GitHub)" <gi...@apache.org> on 2026/01/15 15:55:02 UTC
[PR] KAFKA-19233; Allow fenced members to rejoin streams group with epoch 0 [kafka]
dajac opened a new pull request, #21312:
URL: https://github.com/apache/kafka/pull/21312
This fix allows members to rejoin a streams group with memberEpoch=0
after being fenced, as specified by KIP-848. Previously, the validation
in throwIfStreamsGroupMemberEpochIsInvalid rejected epoch=0 when the
member had a higher epoch on the server.
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] KAFKA-19233; Allow fenced members to rejoin streams group with epoch 0 [kafka]
Posted by "dajac (via GitHub)" <gi...@apache.org>.
dajac commented on code in PR #21312:
URL: https://github.com/apache/kafka/pull/21312#discussion_r2697448082
##########
core/src/test/scala/unit/kafka/server/StreamsGroupHeartbeatRequestTest.scala:
##########
@@ -1003,6 +1003,94 @@ class StreamsGroupHeartbeatRequestTest(cluster: ClusterInstance) extends GroupCo
}
}
+ @ClusterTest
+ def testFencedMemberCanRejoinWithEpochZero(): Unit = {
+ val admin = cluster.admin()
+ val memberId = "test-fenced-rejoin-member"
+ val groupId = "test-fenced-rejoin-group"
+ val topicName = "test-fenced-topic"
+
+ try {
+ TestUtils.createOffsetsTopicWithAdmin(
+ admin = admin,
+ brokers = cluster.brokers.values().asScala.toSeq,
+ controllers = cluster.controllers().values().asScala.toSeq
+ )
+
+ // Create topic first.
+ TestUtils.createTopicWithAdmin(
+ admin = admin,
+ brokers = cluster.brokers.values().asScala.toSeq,
+ controllers = cluster.controllers().values().asScala.toSeq,
+ topic = topicName,
+ numPartitions = 3
+ )
+
+ val topology = createMockTopology(topicName)
+
+ // Join and wait for assignment.
+ var streamsGroupHeartbeatResponse: StreamsGroupHeartbeatResponseData = null
+ TestUtils.waitUntilTrue(() => {
+ streamsGroupHeartbeatResponse = streamsGroupHeartbeat(
+ groupId = groupId,
+ memberId = memberId,
+ rebalanceTimeoutMs = 1000,
+ activeTasks = Option(streamsGroupHeartbeatResponse)
+ .map(r => convertTaskIds(r.activeTasks()))
+ .getOrElse(List.empty),
+ standbyTasks = Option(streamsGroupHeartbeatResponse)
+ .map(r => convertTaskIds(r.standbyTasks()))
+ .getOrElse(List.empty),
+ warmupTasks = Option(streamsGroupHeartbeatResponse)
+ .map(r => convertTaskIds(r.warmupTasks()))
+ .getOrElse(List.empty),
+ topology = topology
+ )
+ streamsGroupHeartbeatResponse.errorCode == Errors.NONE.code() &&
+ streamsGroupHeartbeatResponse.activeTasks() != null &&
+ !streamsGroupHeartbeatResponse.activeTasks().isEmpty
+ }, "Did not get assignment within the timeout period.")
+
+ // Verify we have an epoch and assignment.
+ assertEquals(memberId, streamsGroupHeartbeatResponse.memberId())
+ assertTrue(streamsGroupHeartbeatResponse.memberEpoch() > 0)
Review Comment:
Streams goes to epoch 2 in this case due to the delayed rebalance time. I have updated the assertions to make it clear.
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] KAFKA-19233; Allow fenced members to rejoin streams group with epoch 0 [kafka]
Posted by "dajac (via GitHub)" <gi...@apache.org>.
dajac merged PR #21312:
URL: https://github.com/apache/kafka/pull/21312
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] KAFKA-19233; Allow fenced members to rejoin streams group with epoch 0 [kafka]
Posted by "lianetm (via GitHub)" <gi...@apache.org>.
lianetm commented on code in PR #21312:
URL: https://github.com/apache/kafka/pull/21312#discussion_r2696026999
##########
core/src/test/scala/unit/kafka/server/StreamsGroupHeartbeatRequestTest.scala:
##########
@@ -1003,6 +1003,94 @@ class StreamsGroupHeartbeatRequestTest(cluster: ClusterInstance) extends GroupCo
}
}
+ @ClusterTest
+ def testFencedMemberCanRejoinWithEpochZero(): Unit = {
+ val admin = cluster.admin()
+ val memberId = "test-fenced-rejoin-member"
+ val groupId = "test-fenced-rejoin-group"
+ val topicName = "test-fenced-topic"
+
+ try {
+ TestUtils.createOffsetsTopicWithAdmin(
+ admin = admin,
+ brokers = cluster.brokers.values().asScala.toSeq,
+ controllers = cluster.controllers().values().asScala.toSeq
+ )
+
+ // Create topic first.
+ TestUtils.createTopicWithAdmin(
+ admin = admin,
+ brokers = cluster.brokers.values().asScala.toSeq,
+ controllers = cluster.controllers().values().asScala.toSeq,
+ topic = topicName,
+ numPartitions = 3
+ )
+
+ val topology = createMockTopology(topicName)
+
+ // Join and wait for assignment.
+ var streamsGroupHeartbeatResponse: StreamsGroupHeartbeatResponseData = null
+ TestUtils.waitUntilTrue(() => {
+ streamsGroupHeartbeatResponse = streamsGroupHeartbeat(
+ groupId = groupId,
+ memberId = memberId,
+ rebalanceTimeoutMs = 1000,
+ activeTasks = Option(streamsGroupHeartbeatResponse)
+ .map(r => convertTaskIds(r.activeTasks()))
+ .getOrElse(List.empty),
+ standbyTasks = Option(streamsGroupHeartbeatResponse)
+ .map(r => convertTaskIds(r.standbyTasks()))
+ .getOrElse(List.empty),
+ warmupTasks = Option(streamsGroupHeartbeatResponse)
+ .map(r => convertTaskIds(r.warmupTasks()))
+ .getOrElse(List.empty),
+ topology = topology
+ )
+ streamsGroupHeartbeatResponse.errorCode == Errors.NONE.code() &&
+ streamsGroupHeartbeatResponse.activeTasks() != null &&
+ !streamsGroupHeartbeatResponse.activeTasks().isEmpty
+ }, "Did not get assignment within the timeout period.")
+
+ // Verify we have an epoch and assignment.
+ assertEquals(memberId, streamsGroupHeartbeatResponse.memberId())
+ assertTrue(streamsGroupHeartbeatResponse.memberEpoch() > 0)
Review Comment:
I expect the epoch here is 1 after the single round of join above, correct? If so I think it's not enough to really cover the fix (because we'll be rejoining with an epoch that matches the previous epoch, which is allowed without the fix really).
I expect we need to get fenced minimum at 2 (so previous epoch != 0), to hit the fix. Makes sense?
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org