CPU WMI check returns negative utilization value

[txwomansu]

We check CPU load via WMI.

How is a negative CPU average utilization value possible?

CRITICAL - [Triggered by _AvgCPU<0] - Average CPU Utilisation -3.80% (Over approx 5 sec period)

Thanks!

[mguthrie]

There could either be a logical bug in the plugin, or maybe it could be an issue with the server times being out of sync from each other(?). You can run the following on the Xi server to update the system time.

Code: Select all

yum install ntp
ntpdate pool.ntp.org

[scottwilkerson]

Actually, looking at the plugin developers Change Log it looks like this was fixed in Version 1.44 - 16th June 2011

Sometimes calculated fields based on WMI Raw data which use the WMI field Timestamp_Sys100NS return negative values. This affects checks which were calculated by differencing 2 WMI queries over a period of time eg checkcpu. This is most likely caused by inaccurate timing information provided by the hardware. This most likely only affects virtual machines, probably more on busy host platforms, due to their known inability to return accurate clock information. To counter this, we now check all calculated values and if they are below zero, we make them zero. This is really only a work-around. All calculations like this on VMs will be inaccurate by single digit percentages, especially when checked over short timeframes. We recommended using values for -y and -t as high as you can to improve accuracy. Thanks to Lee Kennedy for notifying us of this problem.

If you update the plugin to any version newer than ( 1.44 ) this should fix this problem