Did you ever solve this problem? I've been using atomikos, spring DMLC and activemq for several years. Recently I started using camel and we are having this exact same problem. With the camel endpoints polling (which effectively doubled the number of message listener containers I previously had) empty queues I am having this same problem where after 12 or so hours the activemq heap blows up (I have mine set at 512MB) and then activemq starts throwing errors and my application crashes and cannot recover (cannot reconnect to the activemq broker).
James Black wrote > > Hi > > ok the problems seems to be that the connection is getting dropped by the > ActiveMQ broker on the server. > > I have now verified that the issue occurs when ActiveMQ runs out of heap > available. > > I have ran two specific tests. > > Note: Although these tests do not allocate that much heap to AMQ we have > seen these issues with much more heap available and believe they would > occur eventually no matter how much heap was allocated. > > === Test 1 === > > Connected our client Tomcat applications to the broker and left them > running overnight. > > I configured the AMQ instance to have only 90MB of heap available to speed > up the process. > With nothing at all occurring on the queues and single topic, the ActiveMQ > installation ran out of heap memory after two hours. > > Again there is nothing on the queues or topic during this time. > > All that is happening is that the Spring DMLC are polling for new > messages. > > This then caused the JMS errors related to Atomikos in the application > logs. > > You can see the GC trying desperately to reclaim memory as the limit is > approached but you can also see a gradual decrease in the amount of memory > that can be GC'ed as time progresses. > > === Test 2 === > > Connected out Tomcat client applications to the broker and ran tests that > pass hundreds of messages per second through the channels. > > Again with a setting of 90MB of heap available to ActiveMQ after a period > of 30 minutes AMQ ran out of memory and we could see that GC was making no > difference. I.e. as in test 1 the memory that could be reclaimed with GC > was decreasing as time went on. > > We are using persistent storage for our queue and single topic. > Our topic is also durable. > > There is very little in the AMQ server logs, occasionally the Out of > memory error will be logged and the disconnection when the heap runs out > can be seen. > > Has anyone seen similar behaviour or have any suggestions as to where to > go next? > > I have application snapshots (VisualVM) of this issue occurring in test 1. > > Thanks to Gary Tully for his help and suggestions previously :> > > Thanks, > > James > -- View this message in context: http://activemq.2283324.n4.nabble.com/Transaction-not-started-XA-problems-tp3565540p4238292.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
