This is always going to be a problem when you have shared locks, and increasing concurrency level will only mitigate the problem, not resolve it altogether. There is, unfortunately, a fundamental problem with a lock-per-fqn scheme, in that the lock needs to exist somewhere, even on a non-existent Fqn (for creates, or removes on nodes that don't exist). We had this problem in PL since locks (per node) were stored in the node, and it involved all sorts of hacks to get it to work, severely degrading performance and still exposing a few race conditions. Not pleasant stuff. Let me think about alternate approaches here, and what can be done. View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4218939#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...