Performance Center Practitioners Forum
cancel

Sitescope using up CPU

Respected Contributor. Soderborg
Respected Contributor.

Sitescope using up CPU

Hi,

I have a problem with Sitescope using up 100% CPU and have >200.000 handles after a couple of days uptime.
When this occur, Sitescope are not able to provide any metrics to Performance Center.

Monitor runs per minute: ~1200
Unix servers monitored: ~100
Windows servers monitored: ~10
Total license points 500; points consumed 480

Sitescope Configuraiton Tool Sizing changed the default config to JVM size to 4096MB, desktop heap size to 8192MB and number of handles to 18,000.

Installation spec:
Windows 2008 sp1 64bit running as a VMware VM (12 GB RAM and 4 cores at 2.30GHz Intel Xeon E5-2670v3)
HP SiteScope 11.33.381 64-bit JVM, Build 290

We have ~10 JMX monitors that often fails (my main suspect for this issue)
There are also several scripts that collects metrics from different Java processeses.

Anyone here with the same issue and perhaps some guidelines on how to solve it?

Thanks

2 REPLIES
Shlomi Nissim
Micro Focus Expert

Re: Sitescope using up CPU

Hi,

I am not an expert in SiteScope so I cannot provide a real solution.

I suggest one (or more of the following)

> Install another SiteScope instance on a different machine and balance the workload

> Change the sampling rate from every 15 seconds to every 1 minute. It has direct impact on Sitescope scalability

Check the sizing calculator - sounds like you already did

> Raise the same question in BAC/BSM forum where I believe SiteScope team members are more active

Regards,

Shlomi

Respected Contributor. Soderborg
Respected Contributor.

Re: Sitescope using up CPU

Hi Shlomi,

Thanks for the reply and suggestions.

I've posted the question in the BSM forum as you suggested.

Changing the sampling rate would not be suitable for us. We need to get metrics often during performance tests. We correlate GC's to CPU-peaks, unexpected peaks in transaction response time to infrastructure issues and so on.

Would be great to be able to either shut down or reuse SSH-connections alternative Sitescope to release handles and threads when not used. (If Sitescope was a system I was running loadtests on, I would susspect a memoryleak :-)

I'll give it a try to spin up another windows server and balance the load. Thanks