You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@activemq.apache.org by "ASF GitHub Bot (Jira)" <ji...@apache.org> on 2026/01/30 13:33:00 UTC
[jira] [Work logged] (AMQ-9852) AdvisoryBrokerTest: Ensure advisory firing occurs within lock to maintain proper ordering
[ https://issues.apache.org/jira/browse/AMQ-9852?focusedWorklogId=1002674&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-1002674 ]
ASF GitHub Bot logged work on AMQ-9852:
---------------------------------------
Author: ASF GitHub Bot
Created on: 30/Jan/26 13:32
Start Date: 30/Jan/26 13:32
Worklog Time Spent: 10m
Work Description: cshannon commented on PR #1641:
URL: https://github.com/apache/activemq/pull/1641#issuecomment-3823774532
I looked at the test and I think I know what is going on but still would need to do some more work to prove it.
The test sends the consumer info for the second consumer (consumer on real topic) to generate the advisory, and then immediately sends the consumer info for the first consumer (consumer on the advisory topic) so it can receive the advisory and get a replay. I believe the issue is that the advisory consumer may receive a duplicate of the consumer coming because it is getting both the "live" advisory and also a "replay" of the advisories because the second consumer is still in flight and being processed as coming online at the same time the advisor consumer is coming online. I'm not quite sure of the order but it seems like that is what is going on where it's just getting 2 copies as the logic for processing the new advisory consumer hasn't quite finished.
When I run the test in a loop in Intellij it usually fails after a few thousand times. If I put in a 10 millisecond sleep immediately after [this](https://github.com/apache/activemq/blob/main/activemq-unit-tests/src/test/java/org/apache/activemq/broker/advisory/AdvisoryBrokerTest.java#L155) line to give the real consumer time to fully come online (which is the goal, so we can test replay) before bringing the advisory consumer online then the test no longer fails.
I'm not sure there's much to do here to fix it other than fix the test, it would be hard to prevent issues in all cases without deadlocks. A duplicate consumer advisory isn't really a big deal as the consumer id is the same so the advisory just gets notified twice but it's the same id so I would expect processes listening to the advisories (like network bridges) to be able to handle it without issue.
Issue Time Tracking
-------------------
Worklog Id: (was: 1002674)
Time Spent: 20m (was: 10m)
> AdvisoryBrokerTest: Ensure advisory firing occurs within lock to maintain proper ordering
> -----------------------------------------------------------------------------------------
>
> Key: AMQ-9852
> URL: https://issues.apache.org/jira/browse/AMQ-9852
> Project: ActiveMQ
> Issue Type: Test
> Reporter: Jean-Louis Monteiro
> Priority: Major
> Time Spent: 20m
> Remaining Estimate: 0h
>
> This popped up with
>
> 2026-01-27T14:10:54.4229847Z [ERROR] AdvisoryBrokerTest>CombinationTestSupport.runBare:107->testConsumerAdvisoriesReplayed:175 ClassCast class org.apache.activemq.command.ConsumerInfo cannot be cast to class org.apache.activemq.command.RemoveInfo (org.apache.activemq.command.ConsumerInfo and org.apache.activemq.command.RemoveInfo are in unnamed module of loader 'app')
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@activemq.apache.org
For additional commands, e-mail: issues-help@activemq.apache.org
For further information, visit: https://activemq.apache.org/contact