Hi - we have a test cluster with three nodes a1, a2, a3. We have configured Tomcat sessions replication with buddyreplication.NextMemberBuddyLocator and one buddy per node. Buddy Replication Question I assume that with this setup we would have something like a2 replicates a1's sessions, a3 replicates a2's sessions and a1 replicates a3's sessions. However what I see (using memory profiling) is that the the memory usage pattern on app1 is virtually the same as non clustered (i.e. a slow drain util major garbage collection). a2 shows a much more rapid depletion of memory and a3 shows depletion at at least a 3x rate of the depletion as of a2. What I was expecting was to see the same memory use on each node and am confused about why I do not. THis seems to imply some basic misunderstanding of how the buddy replication works. Why aren't the memory usage patterns between all three nodes the same? Session Replication Question I am assuming that all replicas of a session are destroyed when the user logs off the node that the source session was created on. Is that the case? The issue we are seeing is that under jmx-console service=TomcatClusteringCache property 'NumberOfAttributes' we see that value always increasing during our load test whereas I was expecting that it would reach a steady state when the number of new logins (session create rate) is matched by the number of users logging out (killing their session). If seems if this number always increases then out of memory is inevitable. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4146670#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...