And you are not fetching in memory state on startup? Odd, there should be no reason for it to lock the cluster up. A few replication events may fail due to timing issues, but what is it that prevents it from starting up then? If you can consistently recreate this, do you have a unit test for it? Some thread dumps of the "stuck" node would be interesting to analyse. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4156763#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...