Killing the perl processes helped me too. A simple 'zmlogswatchctl restart' didn't work; after going to the server status webpage, the service would die with the OP's error message.

$ ps...