I work in an enterprise environment and our monitoring tool is kicking
out tons of tickets on a few servers HighCPUAlert....blah blah.
Upon
inspection I find the CPU usage is quite low and is rarely even
spiking. However, I do find in perfmon that the "System > Processor Queue
Length" is very high and this is tied to our monitoring of the CPU. When I say very high I mean in the 20-100 range with 7 day averages around 25. It should be in the 1-3 range.
This is a windows 2008 EE 64-bit virtual server riding on a ESX3.5 host with 1 CPU allocated. It is running Apache. However, I am also seeing this on win2k3 running IIS with 1 vCPU.
I am trying to wrap my head around this and any help would be appreciated. *The
million dollar question is: How can I determine what application is
causing the thread backup in the CPU and what can I do to resolve it?*
Or, perhaps this is just a mis-read based on a virtual environment?? If that is the case then I still need to be able to document it to convince management that the monitoring needs to be adjusted.
I
have been unable to determine a means for diagnosing this without
scheduling downtime for the server and stopping various processes. I
see in process explorer that Apache launches 254 threads for many of
its httpd.exe processes, but those are using zero cycles and very
little CPU. Furthermore, at least one of the other servers displaying
this behavior is NOT running Apache.
I am happy to post
additional details and/or reports, but I thought I would see what info
would be useful before littering the post with non-useful information.
I appreciate your help as I am a bit under the gun to locate an answer to this issue.