I'm a little concerned about this issue too (RC2). I will need to confirm Monday, but I've seen a memory usage issue using non-durable topics. I don't believe I experienced this issue with the 3/22 SNAPSHOT, which was the most recent snapshot I used prior to RC2 (but I will confirm).
The scenario is: two brokers - A, B publisher connects to A consumer connects to B I have topic and dispatch policies on the brokers such that the queue between brokers is 10000, and the pending queue is 5000. The eviction policy is evict-oldest. I kill off (kill-9) broker B The A broker continues to queue to the B broker queue. This grows to 10000, then the pending queue grows to 5000. This is fine, and continues to report this #, as well as discards increasing. However, I noticed a slow growth in memory. Plus, the journal appeared to grow as well, which really surprised me, since I was sending non-durable messages. Later, I restarted the A broker, and the memory footprint was such that it looked to me like it pulled in the messages from the journal. Again, I will confirm this Monday, because I was actually testing something completely different and didn't notice things such as the journal and restart-footprint until a day later. So, I want to redo some tests with this issue in mind. -- View this message in context: http://www.nabble.com/Out-of-Memory---RC2-t1393354.html#a3811206 Sent from the ActiveMQ - User forum at Nabble.com.
