Hi Guys, First post. Apologies if this is already on here...

Question

I have received the below request as a total newbie attempting to help with our SolarWinds Alerts.

Hi Terry,

The CPU alert on PGDB is giving us lots of alerts for high CPU. The thing with PGDB is that it does spike up to 80-100% CPU quite a bit but never stays there for more than a second or 2 seconds, then it comes back down again. However the spikes are quite regular, maybe 3 or 4 times a minute at very busy periods. What I think is probably happening is that SolarWinds checks when one of these spikes is occurring and then checks 5 minutes later when there happens to be another spike and hence thinks that CPU usage has been high for 5 minutes when it is not really the case.

Is there any way we can prevent these false alerts coming through ?

Thanks

Is there a way we can avoid it polling constantly at the peak periods?

Kind Rgds,

Terry

mesverrum · Accepted Answer

Unfortunately the windows cpu poller is a point in time measurement, as opposed the the way you would see the 5 min running avg in linux.  You can check more often and set your alert trigger condition to require longer waits but unfortunately the metric is going to be what it is.  A couple years ago I tried to script up a rolling average cpu load in powershell to use as a SAM custom script monitor but at the time I didn't know enough to make it what I wanted, and I haven't had time to revisit the topic.  It could be done though.