Details
-
Bug
-
Resolution: Not a Bug
-
Critical
-
None
-
JBoss A-MQ 6.3
-
None
-
User Experience
-
Description
we noticed that our A-MQ container is increasing heap memory usage over time, that after some time hungs the JVM.
Looking at the memory usage, we noticed that "G1 Old Gen" is increasing and not cleaning, even after full GC runs.
In our scenario, we have many client connections (150+), each consuming from many queues (10 per client) and sending to many topics. And one client can be disconnected for a long time (we limited to 3 days, using TTL).
In our logs we see many messages like
2017-11-13 12:13:14,091 | INFO | Q NIO Worker 457 | TransportConnection | 152 - org.apache.activemq.activemq-osgi - 5.11.0.redhat-630254 | The connection to 'null' is taking a long time to shutdown. 2017-11-13 12:13:14,156 | WARN | Q NIO Worker 501 | TransportConnection | 152 - org.apache.activemq.activemq-osgi - 5.11.0.redhat-630254 | Failed to add Connection ID:ABCDEFG-0070-65431-1510336122761-1:497 due to javax.jms.InvalidClientIDException: Broker: myBrokerEFG-0123-broker-abc-01 - Client: ABCDEFG-0070 already connected from null
We also see that the dynamicDestinationProducers javax.management.ObjectName count rises from 168,781 to 508,065 over a period of 16 hours.