I solved part of the issue by changing prefetch size from 1 -> 0. No longer getting the negative queue size.
One thing I am now noticing is that when all of the C/C++ consumers are subscribed to the distributed Queue on one activemq server and from Java I publish the connect request to the Queue on the other activemq server, the throughput is a fraction of what it is when I publish the connect request to the Queue on the same activemq server. Any thoughts on how to performance tune? cart2man wrote: > > I'm having some problems that I hope you can help me sort out. > > Architecture: > 2 activemq 5.2 brokers running on different hosts in a network of brokers > configuration. Each defines a single networkConnector as follows: > <networkConnector uri="multicast://default?group=PROD" > name="default-nc" > networkTTL="2" > dynamicOnly="true" > conduitSubscriptions="true" > decreaseNetworkConsumerPriority="false"> > </networkConnector> > > I have a collection of server processes written in C/C++. Each of these > subscribe to the same queue called "connectrequests" and I rely on > activemq to dispatch messages sent to that queue in a round-robin fashion. > The connect string used here is > failover://(tcp://server1:61516,tcp://server1:61516)?jms.prefetchPolicy.queuePrefetch=1 > > I have a web application written in Java which uses Spring JMSTemplate to > publish messages out to the C/C++ services. The connect string used here > is > discovery:(multicast://default?group=PROD)?initialReconnectDelay=100&jms.useAsyncSend=true > > After a while (~300k msgs), I see the queue size for the connectrequests > queue become a negative number and the C/C++ consumers stop getting new > messages. When I purge the queue using JMX, the messages start flowing > again. > > I've read AMQ-2009 and AMQ-1940 but not sure if this is exactly the same > thing. > > Any ideas? > -- View this message in context: http://www.nabble.com/issues---negative-queue-size--msgs-not-being-delivered...-tp22705295p22710795.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.
