The code really needs to be refactored such that the core install and before/after methods (the callbacks currently implement only the after pattern) are handling atomically and consistently. e.g. In the pseudo code I posted above: 1) An error in one of the callbacks needs to unwind any previously successful callbacks already invoked (not all of them) and then do the main uninstall() 2) If there is an error in the main install(), we don't even want to try to reverse the callbacks since they were never invoked in the first place. You can find an example of this kind of error handling in the MainDeployer/DeployerWrapper. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4039796#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...