Re: [tlug] Process monitoring with Nagios

[Sam Tilders <sam@example.com>]

On Fri, 14 Nov 2003 15:43, Sven Simon wrote:
> Has anybody been trying out Nagios (www.nagios.org) ? I just stumbled
> over their

I've been using nagios to monitor about 450 services on over 70 hosts, 
including linux, windows and routers/terminal servers.

> website and I think this is pretty much what I'm looking for. We have
> rolled out quite
> a bunch of Linux boxes and I need to ensure that some dedicated
> processes run on
> specific machines at specific times. Any opinions on how Nagios handles
> such situations.

Nagios schedules checks to run periodically during specific windows of time. 
If your processes were long running enough you could have nagios check that 
the process was running by syncronising the period for the checks to run.

> Any other tools I should look at? How much of an impact do the agents
> have on the
> systems?

The impact of the nagios checks depends on which checks you're doing. The 
check_procs plugin basically does a ps and grep for the process you're 
checking for. This can time out on systems with very large process lists.
It's possible to write more specific check plugins that may be more optimised 
for finding the process you're checking for, like looking for a PID file then 
checking /proc for the pid, then checking the cmdline in /proc to see if it 
is the correct process. I have had to write some of these to cope with plugin 
timeouts.

Nagios is very good at monitoring services that should always be available, or 
at least available for significant windows of time. I usually have my service 
checks set to every 5 minutes.
However, you can set the units to be seconds and set the check periods much 
shorter. This would of course increase load from the nagios scheduler and the 
execution of the check plugins.

-- 
--
Sam Tilders
sam@example.com
(Move to Jupiter)