Jason918 commented on pull request #12890: URL: https://github.com/apache/pulsar/pull/12890#issuecomment-974591446
> Thanks for the case, I think the root cause is there is an ongoing read op to read data from the managed ledger, but after a consumer closed, some messages will be added to the replay queue, after the previous read op completed, the active consumer might get messages out of order by key. The root cause of this case is `recentlyJoinedConsumers` is cleared in Thread3 when c1 is closed. See here. https://github.com/apache/pulsar/blob/552360455382e33901b6d74c26f883ffb6907ba9/pulsar-broker/src/main/java/org/apache/pulsar/broker/service/persistent/PersistentStickyKeyDispatcherMultipleConsumers.java#L137 So in the following Thread2, when `sendMessagesToConsumers` is called for m10-m14. Previously we are using `recentlyJoinedConsumers` in `getRestrictedMaxEntriesForConsumer` to prevent sending these messages to c2. But now it's cleared. so "maxMessages" is returned instead of "0". See here: https://github.com/apache/pulsar/blob/552360455382e33901b6d74c26f883ffb6907ba9/pulsar-broker/src/main/java/org/apache/pulsar/broker/service/persistent/PersistentStickyKeyDispatcherMultipleConsumers.java#L315 -- 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]
