[cd34]

I wouldn't condone doing a kill for that as you're likely to leave tables in a crashed state requiring a repair or at least a check before you start.

However, with that said, you could do something like:

#!/usr/bin/perl

open F,("</proc/loadavg");
($ld,$junk) = split /\./,<F>;
close F;

if ($ld >= 10) {
  `whateveryouwanttorun`;
}

If the 1 minute load average is above 10, it will run that command. Put that in cron and you should be set. However, I'll say again, and I agree with servhot, I think its a bad idea to kill the server.

Perhaps run something like

mysqladmin -uroot -ppassword processlist | mail username@domain.com

or turn on the slow query log, or do a show processlist in the mysql command line interface or phpmyadmin.

Better to find the cause than to treat the effect.