Hi,
We observe the following bug:
multiply re-entering paging mode on an address with an OpenWire subscriber on
it may make this address not consumable unless either Artemis is restarted or
the operation
retryMessages() is called manually from the Artemis Web-UI. Restarting a
consumer doesn't fix the problem.
For instance, right now, after the test
Consumer Count: 1
countMessages() -> 15405
countDeliveringMessages() -> 0
listMessageCounter() ->
{"destinationName":"test-service#test-channel","destinationSubscription":null,"destinationDurable":true,"count":0,"countDelta":0,"messageCount":15405,"messageCountDelta":15405,"lastAddTimestamp":"01.01.70,
01:00:00","lastAckTimestamp":"01.01.70, 01:00:00","updateTimestamp":"01.01.70,
01:00:00"}
The producer (JMeter) can still send the messages, but the consumer (JBoss 4
with activemq-rar-5.16.5.rar) cannot consume them. After stopping the producer
the message count stays constant unless one of the actions listed above
(Artemis restart or call retryMessages() from the WebUI) is performed.
There are no errors in the artemis.log:
2023-03-13 18:22:27,202 INFO [org.apache.activemq.artemis.core.server]
AMQ222038: Starting paging on address 'applicname/os/test-channel';
size=419821894 bytes (42662 messages); maxSize=-1 bytes (-1 messages);
globalSize=419821894 bytes (42662 messages); globalMaxSize=419430400 bytes (-1
messages);
2023-03-13 18:25:36,863 INFO [org.apache.activemq.artemis.core.server]
AMQ224108: Stopped paging on address 'applicname/os/test-channel'; size=0 bytes
(0 messages); maxSize=-1 bytes (-1 messages); globalSize=0 bytes (0 messages);
globalMaxSize=419430400 bytes (-1 messages);
2023-03-13 18:26:49,828 INFO [org.apache.activemq.artemis.core.server]
AMQ222038: Starting paging on address 'applicname/os/test-channel';
size=419998662 bytes (42680 messages); maxSize=-1 bytes (-1 messages);
globalSize=419998662 bytes (42680 messages); globalMaxSize=419430400 bytes (-1
messages);
2023-03-13 18:29:49,979 INFO [org.apache.activemq.artemis.core.server]
AMQ224108: Stopped paging on address 'applicname/os/test-channel'; size=0 bytes
(0 messages); maxSize=-1 bytes (-1 messages); globalSize=0 bytes (0 messages);
globalMaxSize=419430400 bytes (-1 messages);
2023-03-13 18:31:21,063 INFO [org.apache.activemq.artemis.core.server]
AMQ222038: Starting paging on address 'applicname/os/test-channel';
size=419486088 bytes (42628 messages); maxSize=-1 bytes (-1 messages);
globalSize=419476255 bytes (42627 messages); globalMaxSize=419430400 bytes (-1
messages);
2023-03-13 18:34:32,189 INFO [org.apache.activemq.artemis.core.server]
AMQ224108: Stopped paging on address 'applicname/os/test-channel'; size=0 bytes
(0 messages); maxSize=-1 bytes (-1 messages); globalSize=0 bytes (0 messages);
globalMaxSize=419430400 bytes (-1 messages);
2023-03-13 18:35:53,533 INFO [org.apache.activemq.artemis.core.server]
AMQ222038: Starting paging on address 'applicname/os/test-channel';
size=419536777 bytes (42633 messages); maxSize=-1 bytes (-1 messages);
globalSize=419536777 bytes (42633 messages); globalMaxSize=419430400 bytes (-1
messages);
Testing it with different versions it turned out that the versions 2.21.0,
2.23.1 and 2.24.0 are passing my test (continuously re-entering paging mode).
The versions 2.25.0, 2.26.0, 2.27.0, 2.27.1 2.28.0 fail the test . So the
regression seems to be happened from the release 2.25.0 onwards.
We created a Jira issue ARTEMIS-4095 for it. The header is meanwhile wrong as
the bug is not always triggered by 2nd paging, in the example above it was the
5th, but up to now I never needed more than 8 times entering the paging mode to
trigger the bug.
Please let me know if I can provide more information.
Regards,
Artem
