CPU Critical Threshold

I want to change the timer on CPU threshold. I do not need to get an alert when a CPU hit's 100% for a few seconds one or twice a day.

I would prefer to only get alerted when the CPU is in critical status for a longer amount of time.

Can this be done?

Thanks

Find more posts tagged with

Accepted answers

All comments

bshopp

Yes within the alert definition itself you can specify how long the condition must exist before triggering the alert.

dmtrum

Thank you for the response. The alert is part of a preconfigured/built-in alert, I do not know how (or if you even can) edit those alerts.

xaturbatu

You have to duplicate the alert in order to edit it, there's a button at the top for that.

ewk.jr

I've been trying to figure this out too. I believe what you are asking is not about an alert but the node going red or critical because its cpu is @ 100% for a few seconds.

jholmes

You can duplicate the out of the box alerts and then edit them.

ewk.jr

what does sw use to determine critical or warning? i dont think it has anything to do with alerts but if you look at all the nodes and one may be red because cpu is over the threshold for a short period of time.

dmtrum

You're right. I would prefer if during that few seconds that the CPU is at 100% and the node does go into a critical state, that it does not generate an alert. Alert fatigue is a problem and I am trying to relive some pressure in that regards

stuartd

There are a couple of approaches....

First, as @bshopp suggests you can edit your alert to trigger after the condition has existed for more than a set period of time. That time is up to you to determine. Our CPU trigger condition looks like this:

The box highlighted red is where/how you set the length of time the condition must exist for.

A second option could be to set a custom property on the nodes you know go high, and then exclude them via the "trigger scope" which is the section immediately above the trigger condition. Typically the easiest way to do this is by using a custom property to identify the ones you want to have delayed (or no) alerting. If you go down this road then you would need two (2) alerts to trigger "normally" for most and a second alert to trigger only for certain nodes..

Finally, you could individually change the thresholds for CPU alert levels for any specific node. Typically the thresholds for CPU are set at warning is 80% and critical (aka alert) at 90% use.

dmtrum

Thank you