]
Martin Styk updated WFWIP-23:
-----------------------------
Component/s: Artemis
(was: JMS)
[Artemis 2.x upgrade] Stuck messages in
artemis.internal.sf.my-cluster... queue after restarting nodes in cluster
-----------------------------------------------------------------------------------------------------------------
Key: WFWIP-23
URL:
https://issues.jboss.org/browse/WFWIP-23
Project: WildFly WIP
Issue Type: Bug
Components: Artemis
Reporter: Miroslav Novak
Assignee: Jiri Ondrusek
Priority: Blocker
Labels: activemq, feature-branch-blocker
Attachments: journal-node-2.txt
There are lost messages in scenario where nodes in cluster are cleanly stopped and
started again. This issue was hit with Artemis 2.5.0.Final and WF Jeff's integration
branch WFLY-9407_upgrade_artemis_2.4.0_with_prefix.
Test Scenario:
* start two servers in cluster (JGroups used for discovery)
* send messages to testQueue0 on node-1 and node-2
* wait until consumers on both nodes receive 300 messages
* cleanly shut down 1st and then 2nd server
* leave servers shut down for one minute
* start both servers
* wait until both consumers receive 500 messages
* stop sending messages and receive all remaining messages
Pass Criteria: All send messages are received by consumer
Actual Result: There are lost messages.
Investigation:
There are lost messages which were sent to 2nd node. However they got stuck in queue
{{.artemis.internal.sf.my-cluster.8a7e9e98-2c36-11e8-9737-fa163ea20b26}} during load
balancing to 1st server.
I'm attaching trace logs from client and servers and content of journal from 2nd
server.
This is regression against Artemis 1.5.5 thus setting blocker priority.