I read this document but it gives me no answer. The only section which references MDB is the example section. But this does not explain how an failover works with an MDB. Let me go back to my ealier example (s.a.) You said MDB consume locally. Correct me: i deployed 2 local queue, each on any node. I put message to the queue an node A. If node A crashed the messages hang in this (local) queue. Right? Perhaps my approach is wrong. Goal is following scenario: - any node may produce messages - any node may consume messages - if a node crashes, the other must do the work. Messages should not be left unprocessed in a (local) queue Question: 1. Should the queues be deployed local the any node? 2. Should the queues be deployed as ha-singleton? View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4162743#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...