lucasbru commented on code in PR #21110:
URL: https://github.com/apache/kafka/pull/21110#discussion_r2603991529
##########
clients/src/main/java/org/apache/kafka/clients/consumer/internals/StreamsMembershipManager.java:
##########
@@ -699,17 +710,42 @@ public void
onHeartbeatSuccess(StreamsGroupHeartbeatResponse response) {
processAssignmentReceived(
toTasksAssignment(activeTasks),
toTasksAssignment(standbyTasks),
- toTasksAssignment(warmupTasks)
+ toTasksAssignment(warmupTasks),
+ isGroupReady
);
- } else {
- if (responseData.activeTasks() != null ||
- responseData.standbyTasks() != null ||
- responseData.warmupTasks() != null) {
+ } else if (responseData.activeTasks() != null ||
+ responseData.standbyTasks() != null ||
+ responseData.warmupTasks() != null) {
+
+ throw new IllegalStateException("Invalid response data, task
collections must be all null or all non-null: "
+ + responseData);
+ } else if (isGroupReady != targetAssignment.isGroupReady) {
+ // If the client did not provide a new assignment, but the group
is now ready, update the target
+ // assignment and reconcile it.
+ processAssignmentReceived(
+ targetAssignment.activeTasks,
+ targetAssignment.standbyTasks,
+ targetAssignment.warmupTasks,
+ isGroupReady
+ );
+ }
+ }
- throw new IllegalStateException("Invalid response data, task
collections must be all null or all non-null: "
- + responseData);
+ private boolean
isGroupReady(List<StreamsGroupHeartbeatResponseData.Status> statuses) {
+ if (statuses != null) {
+ for (final StreamsGroupHeartbeatResponseData.Status status :
statuses) {
+ switch
(StreamsGroupHeartbeatResponse.Status.fromCode(status.statusCode())) {
+ case MISSING_SOURCE_TOPICS:
+ case MISSING_INTERNAL_TOPICS:
+ case INCORRECTLY_PARTITIONED_TOPICS:
+ case ASSIGNMENT_DELAYED:
+ return false;
+ default:
+ // continue checking other statuses
+ }
}
}
+ return true;
Review Comment:
We changed it in the last PR so that statuses are always returned...
I think it's better to stick with 'true' here. We can start processing on a
NOT_READY group, we won't have tasks assigned so we won't do anything. On the
other hand, if we prevent starting processing because we incorrectly flag a
group as NOT_READY, it's bad (this can only happen if the broker doesn't have
the last PR).
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
