PCM2.2 high cpu usage problem

[Tony Barrett_2]

Ever since upgrading to PCM+2.2 on the central management station (hooked into NNM 7.01), I've got a 'javaw.exe' process eating up 50% CPU continuously (been like it for days). With the other processes running on the box, CPU is 90-100% 24/7 which is causing polling timeouts in NNM and other slowdowns.

This only happens with PCM+2.2 (2.1 never did this). I'm monitoring some traffic on the switches (but not all yet), but haven't really done much else to the setup yet.

I like PCM2.2, but this high CPU problem is not good.

Any ideas? Are HP aware of this?

[Ralph Bean_2]

Tony -

This could become a very wide-ranging discussion.

My first question to you is: Does your PC meet the system requirements: http://www.hp.com/rnd/products/management/ProCurve_Manager_2.2/specs.htm ?

Note that the system requirements for PCM 2.2 are greater than for PCM 2.1.

If the PC meets the system requirements, is PCM running in a VMWare environment?

Thirdly, what other applications are running on this PC besides PCM?

Fourth, what is the RAM usage profile (that is, amount of swapping)? High CPU use can be a symptom of doing a lot of swapping.

Fifth, is Discovery (nearly) always running or mostly Idle?

What is the rate of events in the Event Browser?

Is packet samping running in Traffic Monitor? If so, how many ports is Traffic Monitor sampling?

Ralph

[Tony Barrett_2]

Ok Ralph, I see what you mean. I guess I didn't expect the h/w requirements of 2.2 to be *that* much greater than 2.1. If it's normal for the javaw.exe process to consume that many CPU cycles, then I guess I'll have to live with it.

The box is a dual cpu 1.8Ghz P4 with 2Gb DDR RAM running Win2k SP4 (not VM'd). The box also runs OV NNM 7.01 (and hooked into PCM). Plently of storage space (RAIDed). RAM usage is ok, with a not excessive amount of swapping. It was just this one process that caught me out. Discovery is (nearly) always idle, and events (inc syslog) are recorded ok (none lost).

I'm monitoring around 60 switches, and intend to have stats/samples collection on around 2000 ports in the end, but I'm currently only monitoring around 200.

As an interim solution, I've enabled hyperthreading on this box, and this seems to have made quite a difference. All '4' processors now run at around 35-40%, and the load is evenly balanced, so I guess PCM (or java) multi-threads pretty well.

I'll keep an eye on things, but prospective PCM+2.2 users should be aware of these issues!

[Ralph Bean_2]

Tony -

The system requirements on the data sheet are intended to cover a wide range of scenarios.

The requirements are stated in more detail in the PCM 2.2 Release Note, which is at:

ftp://ftp.hp.com/pub/networking/software/PCM22-Release-Notes-0607.pdf

on page 5.

Ralph

[Mark Putland]

G'day,

PCM 2.2 is just hopeless with CPU performance we manage 400 switches we are running it on dual zeon with 4g's and it still runs like a dog, looks like we are going back to 2.1 for this reason alone. We have done all the things stated in this forum to increase performance but it makes little difference.

[Matt Hobbs]

What type of Xeons?

My guess for why PCM 2.2 is more resource intensive would be that it performs automatic traffic sampling, versus 2.1 where you had to configure this manually.

Possibly try disabling the traffic monitoring feature in 2.2 to see if this helps the performance.