dajac commented on code in PR #14432:
URL: https://github.com/apache/kafka/pull/14432#discussion_r1363408245


##########
group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupMetadataManager.java:
##########
@@ -983,27 +1052,125 @@ private 
CoordinatorResult<ConsumerGroupHeartbeatResponseData, Record> consumerGr
         return new CoordinatorResult<>(records, response);
     }
 
+    private void replaceStaticMemberInConsumerGroup(
+        String groupId,
+        List<Record> records,
+        ConsumerGroup group,
+        ConsumerGroupMember member,
+        ConsumerGroupMember existingStaticMember,
+        ConsumerGroupMember updatedMember
+    ) {
+        // Write tombstones for the departed static member. Follow the same 
order as the one followed when fencing
+        // a member
+        records.add(newCurrentAssignmentTombstoneRecord(group.groupId(), 
existingStaticMember.memberId()));
+        records.add(newTargetAssignmentTombstoneRecord(group.groupId(), 
existingStaticMember.memberId()));
+        records.add(newMemberSubscriptionTombstoneRecord(group.groupId(), 
existingStaticMember.memberId()));
+        // Cancel all the timers of the departed static member.
+        cancelConsumerGroupSessionTimeout(group.groupId(), 
existingStaticMember.memberId());
+        cancelConsumerGroupRevocationTimeout(group.groupId(), 
existingStaticMember.memberId());
+        // Write a record corresponding to the new member
+        records.add(newMemberSubscriptionRecord(groupId, updatedMember));
+        TargetAssignmentBuilder.TargetAssignmentResult assignmentResult = 
computeTargetAssignment(group, group.groupEpoch(), member, updatedMember);
+        records.addAll(assignmentResult.records());
+        records.add(newCurrentAssignmentRecord(groupId, updatedMember));
+    }
+
+    private TargetAssignmentBuilder.TargetAssignmentResult 
computeTargetAssignment(
+        ConsumerGroup group,
+        int groupEpoch,
+        ConsumerGroupMember member,
+        ConsumerGroupMember updatedMember) {
+        String preferredServerAssignor = group.computePreferredServerAssignor(
+            member,
+            updatedMember
+        ).orElse(defaultAssignor.name());
+
+        String groupId = group.groupId();
+        Map<String, TopicMetadata> subscriptionMetadata = 
group.subscriptionMetadata();
+        String memberId = member.memberId();
+        try {
+            TargetAssignmentBuilder.TargetAssignmentResult assignmentResult =
+                new TargetAssignmentBuilder(groupId, groupEpoch, 
assignors.get(preferredServerAssignor))
+                    .withMembers(group.members())

Review Comment:
   Well, at the moment, a new target assignment will be computed for the new 
static member and that block of code will indeed create the record for it. What 
I meant is that the new static member should actually reuse the target 
assignment of the previous member (vs computing a new one). Then we only need 
to recompute the assignment if there is a group epoch bump.



-- 
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

Reply via email to