Switches, Hubs, and Modems
cancel
Showing results for 
Search instead for 
Did you mean: 

PCM+ brings high CPU load to 2524 switches

SOLVED
Go to solution
Rainer Mittenzwey_2
Regular Advisor

PCM+ brings high CPU load to 2524 switches

Hi together,
i just started to try out thr PCM+, first look ws nice, but actually cpu load on some of our 2524 switches was rising up to 99%!
Exspecially on two of them this happened, so i excluded them from the device list in PCM:
same result, it still rises the cpu load.
I have to end the complete services, to stop this.
Is this a known bug?
Any ideas how to solve it?
Regards
Rainer
6 REPLIES
Matt Hobbs
Honored Contributor

Re: PCM+ brings high CPU load to 2524 switches

I've seen this too. Stopping PCM from traffic sampling the 2500's should hopefully solve your problem. (It did for me)
Rainer Mittenzwey_2
Regular Advisor

Re: PCM+ brings high CPU load to 2524 switches

that is funny,
on some 2500s it works without problems, just 2 or 3 of them have the problem.
Matt Hobbs
Honored Contributor

Re: PCM+ brings high CPU load to 2524 switches

I found that it has to do with how many ports PCM is sampling, and the amount of traffic that is going through the switch.
Steve Britt
Respected Contributor
Solution

Re: PCM+ brings high CPU load to 2524 switches

Rainer,

You can still monitor traffic on these switches without sampling by manually requesting that only statistics monitoring be performed on them. In the PCM Traffic tab, order the ports shown by IP so that the ports for the devices under discussion are all grouped. Then select all of them in a block and right-mouse on the selected group, choose "Manual" from the dropdown menu, and then "Manually enable statistics" from the next dropdown that cascades.

This will ensure that PCM doesn't try to sample on the ports from the 2524s that you're concerned about. Of course you won't see traffic content in your Top Talkers anymore, just the "envelope" of the traffic showing you how much occurred in each Top Talkers category, but it still provides some level of visibility.

Regards,

SVB
Rainer Mittenzwey_2
Regular Advisor

Re: PCM+ brings high CPU load to 2524 switches

Hi,
i found this in help:
The traffic sampling collection utilizes a standard called sFlow (RFC 3176) in newer ProCurve devices; in older devices, HP Extended RMON (XRMON), is employed.
I did as you said: the result ist good, no more cpu stress.
So it seems to me, the sflow is stressing the switch.
We had the phenonema before, when activating sflow on the core switches, with the result, that they suddenly crashed... Disabling helped.

I dont understand it anyway, why this is so stressing...


Rainer Mittenzwey_2
Regular Advisor

Re: PCM+ brings high CPU load to 2524 switches

actually i found this:
2500 supports: XRMON
2600: RMON
4100: RMON

so PCM would not make too much sense in our environment, i guess.

regards
rainer