I've studied the GC root paths for some of these objects, they all look like this: | org.jboss.jms.tx.ClientTransaction | <- value of EDU.oswego.cs.dl.util.concurrent.ConcurrentHashMap$Entry | <- [287] of EDU.oswego.cs.dl.util.concurrent.ConcurrentHashMap$Entry[1025] | <- table of EDU.oswego.cs.dl.util.concurrent.ConcurrentHashMap | <- transactions of org.jboss.jms.tx.ResourceManager | <- resourceManager of org.jboss.jms.client.state.ConnectionState | <- parent of org.jboss.jms.client.state.SessionState | <- parent of org.jboss.jms.client.state.ConsumerState | <- state of org.jboss.jms.client.delegate.ClientConsumerDelegate [Stack Local] | Where the stack local is reported to be on the stack of 'Connection Consumer for dest JBossQueue[...]', which is a thread feeding an MDB. I've collected allocation traces for some of these objects as well, and they all look like this: | org.jboss.resource.connectionmanager.xa.JcaXAResourceWrapper.end(Xid, int) | com.arjuna.ats.internal.jta.transaction.arjunacore.TransactionImple.delistResource(XAResource, int) | org.jboss.resource.adapter.jms.JmsSession.close() | ... earlier frames vary ... | It's strange that the traces cut off there since (looking at the source), no allocations occur in that function. However, I tried to figure out where the trace went from there and my guess is something like MessagingXAResource.end -> MessagingXAResource.unsetCurrentTransactionId -> ResourceManager.createLocalTx, which seems to be a plausible source of these allocations. I don't know where to go from here. Please help :( View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4117935#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...