D Howe <[EMAIL PROTECTED]> writes: > > > I'm trying to setup a network of brokers, as per the documentation, to > provide a level of redundancy. The configuration that I have is one > instance of an application that is producing messages on a queue, a network > of two ActiveMQ brokers running on the same machine and one instance of an > application that is consuming the messages from the queue. > ....
I am pretty much in the same situation right now - testing network of brokers scenarios using AMQ 3.1 and can confirm your observations. My preference right now is to use topics instead of queues even for request/reply scenario. Recovery (reconnect + consumtion of next messages) after consumer shut down seems to work reliable in that case. However this will not redeliver outstanding requests. So requests send between one broker goes down and the clients reconnect to the second broker are lost, but message producer could handle that on application level, because it is not receiving a reply. That's my work around. I am looking forward to the 4.0 release which hopefully fixes that.
