A likely explanation for such a graph is that you're using VM cursors and your consumers are slow. Therefore your messages are piling up in memory. I wonder if producer flow control is enabled at all and why the constant pending message limit strategy is not ditching old messages.
Unless you post your broker XML configuration soon, no one will be able to help you any further. Regards, Raúl. Sent from my iPad On 27 Feb 2013, at 15:40, "mandar.wanpal" <mandar.wan...@gmail.com> wrote: > Hi All, > > Thanks for the replies. From the replies above, I could understand 2 things. > 1. I need to post detailed config. (Christian) > 2. There are cases of AMQ running fine with people. (Gaurav) > > Firstly, configs. > Below are startup jvm params. > /usr/lib/jvm/java-1.6.0-sun.x86_64/bin/java -Xms1g -Xmx1g -Xss256k > -Dorg.apache.activemq.UseDedicatedTaskRunner=false > -Djava.util.logging.config.file=logging.properties > -verbose:gc > -XX:PermSize=128m > -XX:MaxPermSize=128m > -Xloggc:/opt/app/apache-activemq/data/amq_gcstats.log > -Dcom.sun.management.jmxremote.port=1099 > -Dcom.sun.management.jmxremote.authenticate=false > -Dcom.sun.management.jmxremote.ssl=false > -Dcom.sun.management.jmxremote > -Dactivemq.classpath=/opt/app/apache-activemq/conf; > -Dactivemq.home=/opt/app/apache-activemq > -Dactivemq.base=/opt/app/apache-activemq > -Dactivemq.conf=/opt/app/apache-activemq/conf > -Dactivemq.data=/opt/app/apache-activemq/data > -Djava.io.tmpdir=/opt/app/apache-activemq/tmp > -jar /opt/app/apache-activemq/bin/run.jar start > > > @Gaurav: > Yes this system is not a productive system where I am doing load testing and > its not even quadcore. Its dualcore. But, my productive system is also not > that high end because we are at very initial stage in this project and we > may ramp up these systems as per the requirements. I have production system > with dual core 2.67Ghz processor with 8G RAM > From the config I have provided above, If you could help me find out > potential areas where this config can be improved for such setup. > Will it be too much to expect 10 messages/sec with such setup. > > Also main concern I had mentioned was with the increasing FullGC which > ultimately results in AMQ not responding and it switches to its failover > (which also fails at some point). Have you observed any such thing at your > end. I have seen jvm stats via jconsole while these tests. below are the > jconsole stats. > heapusage_amq.jpg > <http://activemq.2283324.n4.nabble.com/file/n4664204/heapusage_amq.jpg> > VMSummary.jpg > <http://activemq.2283324.n4.nabble.com/file/n4664204/VMSummary.jpg> > > I will attach activemq.xml ASAP. > > > > > ----- > Regards, > Mandar Wanpal > Email: mandar.wan...@gmail.com > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/AMQ-halts-and-crashes-after-few-thousand-reqs-tp4663532p4664204.html > Sent from the ActiveMQ - User mailing list archive at Nabble.com.
