On one of my machines, an SBS 2008 box, PCMonitorSrv.exe (2.4.1) takes about 30% CPU, while a SVCHOST.EXE instance that appears to be related to it takes about 20%. The SVCHOST instance stops with the PCMonitorSvc, so I'm assuming they're related.

When the service is started, it runs 0-5% for anywhere from a few minutes to a few hours; then jumps to 50% (total of the 2 services). To my knowledge, this is the only machine (of 10 so far) that does this; at least it's the only one I've noticed. I'm in the process of adding a "Service" Notification on the other 9 to alert me if the service uses > 10% CPU for > 3 minutes to see how widespread it is.

I've restarted the SBS box and uninstalled/reinstalled PC Monitor; no change.

Hello Jeff,

I've been using PCMonitor on two of my servers for three months now and it never used up more than 3,5 percent of my cpu. My server machines are old, xeon 3ghz and windows server 2003 but I never experienced high or medium cpu usage from PCMonitor not even on my personal desktop. Are you sure that this is not isolated on that box? The svchost high cpu usages should increase the cpu usage of pcm but why it's eating away your cpu I don't know. If you uninstall pcm svchost processes have low cpu usages?

Paul

Hi Paul

So far, I'm only aware of it on the one...but I've not spent any time looking at PC Monitor CPU usage in the last couple months. Not until I discovered the problem by accident. That's why I'm setting up a Notification on the other 9 to make PCM tattle on itself, if it occurs.

One can't uninstall the PCM SVCHOST processes per se...SVCHOST is part of Windows. The DLL that SVCHOST is hosting in the instance with the runaway CPU is (likely) installed & uninstalled by PCM. But it could also be a Windows service that PCM is sharing. But I haven't dug into that. What I do know about the CPU-sucking SVCHOST instance is that it stops doing so the instant I terminate PCMonitorSVC, so it's reasonable to assume PCM is using that SVCHOST instance.

If it's just the one box, I'm hoping I can isolate it to one particular monitor so I can turn off that monitor until MMSoft can fix it (if it's, in fact, a PCM bug). Sad thing is that on an SBS box, there's a LOT of stuff being monitored, so it could take a while.

Can you please export the configuration and email it to us (without the account info)? We will investigate this.

Thanks, Marius, will send it shortly.

Update: My new Services Notification for PCM Service > 10% for > 3 minutes has fired twice on my other company's system (non-SBS, all WS2008 R2).

When I logged on to one of the problem machines, PCM Monitor Service was at 63% 30 minutes after the notification fired. It has since dropped back to 0-2%.

On the SBS machine I mentionned in my OP, PCM was about 30%. These VMs, however, have only 2 CPUs, where the SBS box has 4. Adjusted for that, the CPU usage is the same.

On the other high-usage box, PCM was back down to 0-2% by the time I logged on. But by that time it was an hour since the event fired.

These are a completely different domain from the SBS 2008 box, FWIW. Both machines are WS2008 R2 Core, dedicated DCs. Their only obvious commonality with the SBS box is that they are DCs. Not even using the same AV, and there isn't much beyond AV that's installed on these two.

Marius, I'll send you .reg files on these as well, shortly.

PCM CPU usage monitoring has not been active on all servers on this other domain until a couple hours ago due to my failure to accurately reverse-engineer the Registry settings I'd have to hack in order to push the settings. However, the 2 DCs discussed in this post had the latest Notification reg entries because they were set manually.

Config settings are now resolved on the other machines; hopefully we won't have more in the next 24 hours. But I'll be sure to let y'all know!

Hope it gets fixed, but...

Moral of the story: PCM needs to monitor itself with EVERY possible relevant monitor and report problems. That should not be an option, or a user configuration. If the user opts to monitor NOTHING, PCM should be monitored.

And that makes it even more imperitive that the "Computer is up an running" notification be replaced with a "PC Monitor Service is launched" notification. PCM will NEVER be able to accurately report ANYTHING at the moment the service is launched.

Thank you Jeff,

We will look into this ASAP - PC Monitor (when idle with no commands being sent) should not perform any tasks except monitoring the parameters specified in the notifications.

Therefore it should not be anywhere near 60% CPU - I will look at the config you've sent - thank you for that.

We monitor around 20 test machines (including VM's) and not being able to reproduce this just yet but we will investigate with all sorts of configurations.

We were investigating this for 2 days now and we cannot reproduce it on any test servers.

There is a new beta 4 update for the PC agent available on the downloads section on the forum - please try to see if you can replicate the issue with that version. We've changed the priority of all threads especially the performance counters reading thread. As well the reading interval has been changed.

The latest 2.5 beta 5 improves the event watch handling - this should fix the issue you are experiencing. Thanks Jeff for all your help.

Thanks, Marius, you fixed it with 2.5 beta 5!

I really appreciate your responsiveness. Looking forward to future PC Monitor improvements.