We are implementing BMC Patrol to monitor Windows and Linux servers in our environment (using it to monitor and report on performance stats, uptime, event logs, etc). Since Linux allocates "free" memory to things like cache and tempsfs, the "Memory Free" element can sometimes show 1% or less. Right now we have set our yellow Warning threshold on Memory Free to 1%. My question is, is it worthwhile to monitor free memory at all? To me , getting alerted to low memory condititions in a Linux box is not really helpful. How do you out there with large Linux server farms monitor free memory and alert on it?

I don't have a "large" server farm, but I use Nagios and it's currently monitoring 16 Linux hosts. I have to admit, though, the only machines I'm checking free memory on are HP Proliant servers that allow SNMP checks (using the HP Proliant Support Pack).

For everything else, I just check the swap levels - high swap == low memory.

Take a look at the line in the 'free' command that starts with "-/+ buffers/cache". The 2nd value (under the free column) is a better indication of how much memory the system could potentially consume.