Comment 2 for bug 620441

Scott, I've experienced problems with SIGKILL's corrupting mysql databases before, so the workaround given might be a little misleading. MySQL can actually be stuck waiting on a single query to finish before it can shut down, and some queries might take *hours*.

That said, its a known tradeoff if you're using MyISAM that you may corrupt tables. The workaround is to do as suggested above, and also move to transactional storage such as InnoDB. With InnoDB the database integrity will survive a SIGKILL (though the next startup may take several hours in recovery if the system has large transaction logs).

5 minutes seems an appropriate length of time to wait. However, is there any way to cancel it?