congbo

@tjiuming yes PendingAck is TP

hi @TakaHiR07 , sorry it took a long time to reply. https://github.com/apache/pulsar/pull/17051 This Pr will fix the `reader.hasMoreEvents` has been blocked problem. please re-build the branch-2.9 or wait for tag...

@TakaHiR07 Hi what version of the client you are using? ``` 18:27:19.260 [transaction_coordinator_TransactionCoordinatorID(id=6)thread_factory-70-1] ERROR org.apache.pulsar.broker.TransactionMetadataStoreService - End transaction fail! TxnId : (6,2309004), TxnAction : 1 org.apache.pulsar.transaction.coordinator.exceptions.CoordinatorException$CoordinatorNotFoundException: Transaction coordinator with id...

@Technoboy- transaction ack command will carry the batchSize, if the batchSize is not in pendingAcks(the message redeliver hasn't been sent to any consumer), we will not get the correct batch...

> @congbobo184 Would it be possible to describe the impact of the bug that was fixed by this PR? The current description describes implementation level details of the issue. However,...

@Anonymitaet @codelipenghui I have fixed the comments, please review again. Thanks! :)

@codelipenghui because, if check the epoch before polling the message from the receiver queue, the user invoke `redeliverUnacknowledgedMessages` and the client checks the message epoch at the same time, the...

@shibd yes, if the user use multi-thread call receive and redeliverUnacknowledgedMessages, we also can receive old epoch messages. in the most cast, use redeliverUnacknowledgedMessages for cumulative ack, so the user...

@shibd > > We should clean up the receiver queue when users invoke the redeliverUnacknowledgedMessages method? > > > > We can make them happen on a single thread. And...

This test has not failed for a long time, it may have been fixed, so close this issue, if it occurs, can reopen it