It looks like I over did it with introducing write lock on MainDeployerImpl::process (which now creates deployment contexts). The deadlock can be seen here: - https://jira.jboss.org/jira/browse/JBAS-6529 - http://lists.jboss.org/pipermail/jboss-development/2009-February/013693.html My idea was to let others have read lock - the one's that are adding/removing on already concurrent maps, while the one's that actually process things should have a write lock. Is this really good/necessary? Or what would be the right way to have these locks? I used to have just write lock on shutdown. Which I guess could result in the same lock Brian sees if he would apply similar construct on the shutdown? View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4211713#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...