Showing results for 
Search instead for 
Did you mean: 
Create Post
Level 8

Wrong CPU Percent Load Calculation via Linux Agent

Hello THWACK members, I have a little misunderstanding on CPU Percent Load calculation.

I have this following issue with one of my environment's machine on CentOS, polling through Agent and default Orion CPU & Memory poller. Top CPUs by Percent Load shows what it has 100% peaks from time to time, but in reality if I used htop, I see what load average is very high and system is overloaded, but CPU in Percentage not increased the value of 60%, Orion displays 100% Utilization. 




How Orion default CPU & Memory Poller calculating the value of percentage loading on Linux systems?

0 Kudos
2 Replies
Level 18

Linux CPU calculations (along with RAM calculations) are a many and splendorous thing - meaning the way you (or a particular utility, like top, htop, gtop, and more) calculate it has as much to do with how you feel about CPU/RAM as it does about the actual hardware.

It's also why you can use Orion pollers (not to be confused with a Universal Device Poller, or UnDP), which you get to by going to Settings, All Settings, Manage Pollers. At that point, you can specify which values (OID) and which calculations (the math) you want to use to arrive at YOUR perfect CPU utilization. Then you can assign the new poller to a specific machine or group of machines (based on machine type, vendor, custom property, etc) and that will be used in place of the built-in CPU utilization.

You can find more details on which OIDS and calculations you might want to use here:

And this article on the Customer Success Center should help you get started with Device Pollers:

Leon Adato | Head Geek
"Measure what is measurable,
and make measurable what is not so." - Gallileo

0 Kudos

@adatole Thanks to your reply.

It's a good advice, but I'm more interesting about how default calculation works. Like you provided the explanation of RAM consume. Use another poller not completely good to this server and some part of our environment because not all server admins configure snmp.

0 Kudos