Capacity planing
A note about memory consumption of MySQL and Apache. Both are more or less deterministic and you ought to do some capacity planning beforehand, not just restarting the server every time it couldn't survive a load spike because of misconfiguration. In case of MySQL sum up all allocated buffers, caches, threads cache (number of ready-to-serve threads) and their stack size, etc and you get the number. MySQLTuner can do it for you -- the line looks something like (besides other important metrics):
[OK] Maximum possible memory usage: 352.0M (48% of installed RAM)
In case of Apache it's harder, but empirical way should be the way to go. Having limited number of workers in Apache configuration, you just need to estimate an average memory consumption per worker -- look at ps, top, etc. Then multiply the value by the number of workers. If you run something like mod_php then there's also a way to restrict memory usage at the side of your runtime.
Monit side
Monit can restart a service (e.g. MySQL) when it matches a certain limit (e.g. consumed memory). But it will be a mistake to do in most cases, because MySQL maintains buffers and cache and serves clients. Restarting would cause performance degradation and service outage. Use it only when you can't restrict memory consumption in any other way. For example:
check process mysql with pidfile /var/run/mysqld/mysqld.pid
start program = "/etc/init.d/mysql start"
stop program = "/etc/init.d/mysql stop"
if totalmem > 256 MB for 10 cycles then restart
group database
It's also possible to monitor system-wide memory consumption, but be aware that Linux memory model is complex and it's really hard to tell what amount of memory a process really consumes. System-wide metrics like free memory can also be misguiding because values of disk cache (cached memory), shared library memory, and many other aspects should be taken into account.
Here's example from Monit's default Debain default config:
check system myhost.mydomain.tld
if loadavg (1min) > 4 then alert
if loadavg (5min) > 2 then alert
if memory usage > 75% then alert
if cpu usage (user) > 70% then alert
if cpu usage (system) > 30% then alert
if cpu usage (wait) > 20% then alert
You can use exec action (then <ACTION>) to restart your server, but I really don't recommend you to follow this path, but rather do some capacity planning.
