Hi ! I just finished a test with one fixed queue for the answer from the server, instead of using temporary queues.
On the client, for every message the client want to send, I associate a JMSCorrelationID, and I create one consumer on this fixed queue with a selector on this CorrelationID. The server receives the message on one of the three queues, and send an answer with the CorrelationID wait by the client consumer. First, It's 10 times slower than using a temporary queue for each thread. It's the same problem. The server session which receives 70% of the trafic is the first session to stop sending answer on the fixed queue. And it is just after sending 1000 answers with the same session. Eric -- View this message in context: http://www.nabble.com/Prefetch-Limit-of-1000-....-without-acknowledges-tp15180095s2354p15206165.html Sent from the ActiveMQ - User mailing list archive at Nabble.com.