New Relic Server Monitor causing high WMI CPU usage

Question

I'm using New Relic Server Monitor to monitor our servers. Unfortunately, starting the New Relic service causes WMI to use 25% CPU. Stopping the New Relic service causes WMI to drop back down.

I've course, I've reinstalled the Server Monitor, but that didn't help.

I've tried to restart the WMI service. But, the system warns me that I'd be stopping four other services. These are Amazon servers, so things like EC2 console and Citrix client are running. I didn't restart WMI because I don't want to be locked out of my server or something.

Is there anything else to try? If I can't resolve this, I'll have to stop using New Relic, which would be bad because it's been so good for us.

Answer 1

We tried installing a number of hotfixes for WMI from Microsoft, but none of them resolved this issue. This was only fixed for a number of our Windows 2008 R2 servers when we installed the Windows Server 3.2.6.0 version of the agent (https://docs.newrelic.com/docs/release-notes/server-release-notes/windows-server-release-notes/windows-server-3260). On some servers CPU usage dropped from an average of 20-40% for WmiPrvSe sessions down to a much more reasonable 1-3%.

Answer 2

This is most likely caused by a bug in unpatched WMI. Microsoft has released a bunch of hotfixes for high CPU/memory usage of WmiPrvSe.exe. There are a bunch of hotfixes for each OS so rather than list them all here we recommend googling "WmiPrvSe.exe CPU usage Windows" here is a link with the appropriate hotfix collection or MSDN article. An example for Windows Server 2008 R2 can be found here: http://social.technet.microsoft.com/wiki/contents/articles/1398.list-of-wmi-related-hotfixes-for-windows-7-and-windows-server-2008-r2.aspx