Hi Damien, Is there any chance that you built the JMS component from SVN this week? I introduced a bug that might have caused something like this and it was exposed in SVN for a few days (its fixed as of Thursday or Friday). If that isn't the case, could you let us know exactly what version you are using and possibly post some of your configuration?
Thanks, Chris -- Chris Custine FUSESource :: http://fusesource.com My Blog :: http://blog.organicelement.com Apache ServiceMix :: http://servicemix.apache.org Apache Directory Server :: http://directory.apache.org On Sun, Apr 5, 2009 at 10:37 AM, dmorcellet <[email protected]> wrote: > > I think there is a memory retention problem in the JMS binding component. > > Here is my test scenario. > > I have a JMS provider service unit and a JMS consumer service unit linked > together. > I have a simple Java program that continuously sends text messages to a JMS > topic (managed by the servicemix embedded activemq server). My consumer > listens on this topic and sends the received message to the JMS provider. > I have another simple Java program that listens to another JMS topic (the > one the provider is writing to). > The consumer endpoint uses a in only MEP (in a first attempt, I used an > in/out MEP, but I rapidly realized that messages were accumulating inside > ther servicemix/activemq server). > > I use 2 message sizes. > I'm using jconsole to monitor the memory consumption of servicemix (and the > state of my topics through JMX beans). > With short messages (20Kb xml text), I have a 200Mb memory retention after > sending about 1 million messages. > This week-end, I'm running another test with 1Mb XML messages (I will have > a > result tomorrow morning). In that case, there are fewer messages that pass > through servicemix. > > Using visualvm (to count instances), I found a map > (JMSComponent.knownExchanges) that contains roughtly as many entries as the > number of messages who did pass through servicemix. This map contains > message IDs. > > I think the memory retention problem is proportional to the number of > messages exchanged. With bigger messages, the time to reach 200Mb memory > leak will be far longer than with 20Kb messages. > > Thanks in advance for any answer / correction / patch / workaround. > Best regards, > Damien MORCELLET > > PS : I'm @home and I do not have the source code for this test (xbean.xml > files, Maven projects and test Java emitter/receiver). > > -- > View this message in context: > http://www.nabble.com/Memory-leak-when-using-the-JMS-binding-component-in-version-3.3-tp22895746p22895746.html > Sent from the ServiceMix - User mailing list archive at Nabble.com. > >
