anvinjain opened a new issue #9540:
URL: https://github.com/apache/pulsar/issues/9540
**Describe the bug**
Pulsar client does not shutdown (deadlock occurs internally) when DLQ is
configured in the consumer and failures are being produced to DLQ topic. This
happens intermittently. I have been able to reproduce locally more often than
not.
**To Reproduce**
Steps to reproduce the behavior:
1. Produce messages to a topic and build a large backlog (tens of thousands
of messages to ensure consumer is occupied for some time)
2. Run a consumer with sub-type=SHARED and deadletter policy configured
(redeliver-count = 1)
3. Intentionally fail messages (negative-ack) so that they would be produced
to DLQ
3. Shut down consumer after 30 seconds or so by calling consumer.close
followed by pulsarClient.close.
4. main will not exit because consumer.close will block indefinitely
**Expected behavior**
Expected program to exit cleanly once consumer and client are shutdown.
**Desktop (please complete the following information):**
- OS: Mac OSX 10.14.6
- Pulsar Standalone 2.7.0
**Additional context**
1. Consumer code to reproduce and thread dump of a stuck consumer:
https://gist.github.com/anvinjain/9799d15b9c0edaf363ddfe5067301517
2. Producer code not added in the gist. Trivial, random messages in a
partitioned topic (partitions = 4)
3. If you look at thread dump, you can see that unackedmessagetracker.stop
is waiting on writelock which is acquired by the periodic timer task that runs
in UnackedMessageTracker. The timer task in turn is blocked on producer.send
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
