"resolved" for today but you will be back at some point asking why this process is so horribly slow. At that point we will ask you again why you didn't just use referential integrity and cascading deletes instead of recreating a feature that is already available (and MUCH faster because it doesn't use cursors). You really should try to handle this with proper architecture instead of throwing more code at the problem. And by all means don't use a cursor for deletes. This looks like a generic delete procedure that will be called every time you want to delete something. Just my 2¢.