"galder.zamarreno(a)jboss.com" wrote : | Looks to me the different caches accessed by your clients are not clustered which is the root of your problem. If they're clustered, modifications could be replicated to other cache instances or modifications could result in invalidation messages to other nodes. Either of these two options would allow other clients to access the most up to date data. In the case of replication because the cache contains up to date data. In the case of invalidation because after data was updated in another node, the cache would not contain the data and it will result on the cache store being queried. Yes. That's exactly right. In the interest of keeping our solution as simple as possible, we have chosen not to use clustering right now. View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4239125#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...