We do apply session affinity. Look at this way: 1. Cache A starts 2. Cache A adds 10 nodes to the cache 3. Cache B starts 4. Cache B 'gets' the 10 nodes thus causing a gravitation After #4 in the sequence we end up with the weird buddy rep settings as discussed above. This is exactly what we are doing in the test provided. We only go through the sequence once, we do not keep modifying/adding etc to the cache. (There is also a 5 second sleep between 3 and 4). Just to clarify: We do not add data to the cache while trying to read it. We do not modify data to the cache while trying to read it. In fact there is no concurrent access at all in the test provided. The issue is clearly related to data gravitation, in the logs we see that the buddy backup is correctly setup before gravitation occurs (i.e. after #3 in the sequence above), but has changed to include itself after data gravitation has occurred (after #4 in the sequence). I'll try and make a comparison run with 2.0.0 GA and post the logs, if behavior is unchanged then we should get the same buddy backup result. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4117065#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...