Too many expired queue messages on intermittent consumer causes ActiveMQ to hang
--------------------------------------------------------------------------------
Key: AMQ-1787
URL: https://issues.apache.org/activemq/browse/AMQ-1787
Project: ActiveMQ
Issue Type: Bug
Components: Broker
Affects Versions: 5.1.0
Environment: Linux
Sun JDK 1.6.0
Reporter: Brian Desai
I have a case where a queue consumer may become inactive for periods of time.
If too may messages are sent to this queue, and they eventually expire (setting
the TTL) before the consumer comes back online, then once the consumer does
come online, I see many of the following messages printed to the broker console.
{quote}Failed to fire message expired advisory{quote}
Once this happens, it seems that no more messages go to any queues, not just
the queue whose consumer was down for a while. I've looked at the web console,
both before and after this occurs, and the message counts for all queues stop.
It appears that the broker just gets hung. I have to kill the broker, and wipe
out the data storage in order to get back to a working state.
I've tried various memory settings, with the same result. My TTL was fairly
low (60 seconds), as I was testing out the ability to expire messages. The
number of messages in the queue that were expired were in the 1000+ range.
With a low number of expired messages (< 100), there was no problem.
As a work around, I can disable sending of the advisory messages, and turn off
processing of expired messages, but this is not ideal.
{code:xml}
<beans>
....
<broker ... advisorySupport="false">
<destinationPolicy>
<policyMap>
<policyEntries>
<policyEntry queue=">">
<deadLetterStrategy>
<sharedDeadLetterStrategy processExpired="false"/>
</deadLetterStrategy>
</policyEntry>
...
</policyEntries>
</policyMap>
</destinationPolicy>
...
</broker>
</beans>
{code}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
