Cisco 3850 & 3650 switches showing incorrect CPU utilisation

Question

Hello all,

NCM is reporting lower CPU for our 3850 & 3650 switches.  Eg. for one particular 3850 stack, it shows 70%, whereas the switch CLI shows;

#sh process cpu sorted

Core 0: CPU utilization for five seconds: 99%; one minute: 98%;  five minutes: 98%

Core 1: CPU utilization for five seconds: 99%; one minute: 98%;  five minutes: 98%

Core 2: CPU utilization for five seconds: 97%; one minute: 98%;  five minutes: 98%

Core 3: CPU utilization for five seconds: 98%; one minute: 98%;  five minutes: 98%

The objectve is for NCM to display the 5 min average across the 4 CPUs.

We are running NCM v7.5.1.  I logged a case with Solarwinds and after some investigation with SNMP Walks they suggested I submit a feature request.  Can anyone assist with a workaround please?  Perhaps a custom SQL node alert or poller can resolve this?

Many thanks!

bdufresne · Answer

So your question is specifically about individual CPUs on devices within the stack, and the Average CPU load gauge from the screenshot above is an average of all CPUs from all switches in the stack, correct?

If that is the case, you may need to add a custom resource to the node page that looks at individual CPUs, or perhaps even a table.  As far as a proper alert, it would look into the Alert Lab: https://thwack.solarwinds.com/community/labs_tht/alert-lab

netw0rks · Answer

Following on from the above, it's clear that the chart shows the average CPU across all the switches - which is correct at 70%.  However we need an alert to trigger if any single switch CPU goes above 85%.  I've tried to set this up but it didn't work, if someone can give me some pointers on this it would be much appreciated.

Thank you.

netw0rks · Answer

Thank you for your response.  Can the moderators move this to the NPM section please?

In response to your questions;

Preferred Cisco MIB: CISCO-ENVMON-MIB

Device template: Cisco IOS

OID: Where can I check this?

SNMP Polling Interval: 120s

Btw the Switch Stack section in the node details shows the correct average CPU percentage per switch in the stack;

However the CPU chart in the Summary section shows 75%;

On checking the Switch CLI it's been 98% approx for the last 72 hours.  Since the chart shows 75% our alerts are not being triggered as these are set to 85%.

Thanks