Tool to diagnose Total Percentage Interrupt Time is too high
Hi, The product knowledge has a link to a 2004 version of Kernrate Viewer without any indication of support for 2008 R2. Any suggestions of what else I can use to diagnose the problem. It is a vmware environment but I can't easily override this one. Thanks, Rudi
July 25th, 2011 5:18pm

% interrupt time for vm's is a pretty useless counter threshold. instead of looking at a percentage, look at the interrupts/sec. you probably find that the interrupts/sec are lower than for a physical server. Bc the cpu scales with vm's it's going over the threshold (over threshold when the vm is doing nothing, with load its under threshold as it has cpu power).Rob Korving http://jama00.wordpress.com/
Free Windows Admin Tool Kit Click here and download it now
July 25th, 2011 5:44pm

Thanks for the feedback, but I don't understand how cpu scaling is affecting the performance counters. Doesn't the guest thinks it has, for instance, a 2.4 GHz processor all the time. If interrupt percentage and processor percentage are calculated inside the client, I'm not sure how the host allocates resources impacts the counters unless there's contention. rudi
July 28th, 2011 12:20pm

just open perfmon and check interrupt/sec yourself... or set the minimum cpu speed for your vm's to like 400mhz, you probably never see this alert again. Rob Korving http://jama00.wordpress.com/
Free Windows Admin Tool Kit Click here and download it now
July 29th, 2011 2:33am

I don't understand how that helps. If the consumed is 1 Ghz and the active is 2 Ghz, setting a reservation to 400Mhz doesn't do anything until there is contention for CPU. The host has lots of cpu to share. Rudi
July 29th, 2011 3:26pm

all i know is that in 99,99% of the time %time counters in vm's don't work as the clock is dynamic so 1 interrupt takes a lot longer than when the cpu is at full capacity. Again, you can easily check this by starting up perfmon and log interrupts/sec.Rob Korving http://jama00.wordpress.com/
Free Windows Admin Tool Kit Click here and download it now
July 30th, 2011 3:48am

Hi, I'm not sure why this was closed. Thanks for the info. Is there a good source for the explanation? I have heard this answer for other monitors on a VMware VM but I don't have a good authority on the explanation which would satisfy a change management request. I have looked at the counters and it's reporting properly. What am I supposed to learn from the interrupts/sec? Thanks, Rudi
August 2nd, 2011 3:09pm

There is a lot of developer turds hanging around in the perf counter space on any piece of software. Not all of it was ever meant to mean anything concrete. So the example of %interrupt time is a good example, as is %CPu - these are not useful for proving anything is good or bad, ugly or pretty. They are just pieces of data kind of like the RPM of a car. You can tell if it is started because the value bounces around a bit. But having an RPM of 2000 doesn't mean anything other than the engine is running at 2000 RPM. You don't make decisions at these levels. Whey you say you want to diagnose the %interrupt counter - what do you mean. Do you want to understand what it means and whether you should pay attention to it (the other replies in this thread talk to that question). Do you actually have a problem and you are just hiding what it is from the other readers in this thead and talking about this other counter because you feel there may be some kind of predictive or counter-predictive relationship between your problem and this particular counter? "Never look at %CPU for more than 10 seconds. You'll go blind" Microsoft Corporation
Free Windows Admin Tool Kit Click here and download it now
August 2nd, 2011 3:21pm

Hi Dan, I want to increase the threshold but I need to explain my recommendation.
August 2nd, 2011 5:51pm

If it were me, I'd start with the reasoning "it alerts too often, so we eitehr shut it off or figure out a better threshold"Microsoft Corporation
Free Windows Admin Tool Kit Click here and download it now
August 2nd, 2011 6:34pm

What am I supposed to learn from the interrupts/sec? Thanks, Rudi Did you try anything i've said? 1. when the vm is doing nothing, the interrupts/sec is low, but % interrupt TIME is high 2. when the vm has load the interrupts/sec is higher (usually still a lot lower than physical servers), but % interrupt TIME is low. Reason the amount of cpu power given to the vm. i've just explained this a few times, but ignore it if you want. For vm's there's only 1 good way to go. Disable the % interrupt time. If you want to monitor this (imho there are way better perf indicators for vm's), find out the "normal" interrupts/sec and create a threshold for that counter. Rob Korving http://jama00.wordpress.com/
August 3rd, 2011 7:20am

Hi, I am seeing this on one machine but not another. That's a bit of progress. I will mark as answer. Thanks for your help.
Free Windows Admin Tool Kit Click here and download it now
August 3rd, 2011 10:50am

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics