[Nagiosplug-help] check_cpu with SNMP

[Ingo Lantschner]

Hello,
before reinventing the wheel I rather start a discussion and hope for  
input:

Plugins like check_snmp_load.pl check for the CPU-usage in percent by  
quering the OID ssCpuIdle (somewhere below 1.3.6.1.4.1.2021.11). This  
counter is not well implemented on all systems (sometimes it shows  
the average since uptime, f.e. on Ubuntu Server and probably on all  
debian-systems) and therefore deprecated. All recommendations state  
to use ssCpuRawIdle instead. Since this is a counter, you have  
calculate the usage yourself. I see different possibilities for this:

1. Write a statusfile to the disk of the Nagiosserver (where the  
plugin resides): Not recommended following the guidelines
2. Poll twice, with a 10 seconds delay - this makes the plugin stay  
in the queue for at least 10 sec. doing nothing: Possible, but what  
if there are 1000s of such checks? And because of the small delta,  
network-latencies can get into the calculation.
3. Write the counter as it is to a local rrd (or other database) and  
poll this rrd by a special plugin: requires rrdtool installed and  
configured.

I would prefer solution 3, since I am interested in recording the  
performance data anyway. So there is no overhead in polling and  
writing. But may be someone already has a better solution for this.

Comments are welcome, cheers, Ingo.

-- 
Ingo Lantschner
Vienna/Austria
Mob (+43-664) 143 84 18