Very High CPU Utilisations

[petenixon]

I have a few production switches (x2 multilayer, x1 layer 3) that have minimal CPU usage (at around 20%) but entering a show run will spike the CPU upwards of 90%. This is something i've not noticed before (anywhere), and was only noticed because I was investigating high CPU utilisations on the L3 switch. Is it normal for a switch to do this? The 2 multilayers have a moderately sized config and the L3 config is quite large, but contrasting this to a comparable sized text file on my workstation it only jumps the CPU to around 30%.

Thanks.           

[Edison Ortiz]

Model and IOS version?

[petenixon]

Multilayer is 3750x, IOS 15.1, L3 is a Catalyst 4000 and the IOS is 12.2.

Thanks.

[Edison Ortiz]

Please post the 'show process cpu' output before and during the high CPU event.

[petenixon]

The correct IOS for each are:

Multilayer: 12.2(53r)SER

L3: 12.2(33)SXH5

Before running a show run (on the L3):

CPU utilization for five seconds: 17%/9%; one minute: 9%; five minutes: 8%

PID Runtime(ms)   Invoked      uSecs   5Sec   1Min   5Min TTY Process

303  27007170603414438445          0  3.59%  1.63%  1.59%   0 Port manager per

   8   9417002681831341989        514  1.03%  1.32%  1.00%   0 ARP Input

146   5074204803233069416          0  0.71%  0.64%  0.61%   0 IP Input

   5   402347664  25984773      15483  0.71%  0.29%  0.28%   0 Check heaps

  20   5651730643014806819          0  0.71%  0.25%  0.19%   0 IPC Seat Manager

199   287328804 205959437       1395  0.47%  0.42%  0.40%   0 CEF: IPv4 proces

153    49314648 120238459        410  0.15%  0.08%  0.06%   0 ARP HA

148   1485301921471432108        100  0.15%  0.12%  0.10%   0 ADJ resolve proc

151         168       141       1191  0.07%  0.11%  0.04%   3 Virtual Exec

192   112436416  52997820       2121  0.07%  0.09%  0.08%   0 IPC LC Message H

167    31255008  87977107        355  0.07%  0.07%  0.06%   0 DHCPD Receive

And after:

CPU utilization for five seconds: 71%/18%; one minute: 18%; five minutes: 10%

PID Runtime(ms)   Invoked      uSecs   5Sec   1Min   5Min TTY Process

151        5640      4482       1258 46.00%  7.44%  1.61%   3 Virtual Exec

303  27007177403414442503          0  4.31%  1.66%  1.59%   0 Port manager per

   8   9417005081831342523        514  1.75%  1.14%  0.98%   0 ARP Input

146   5074207683233071038          0  0.55%  0.66%  0.62%   0 IP Input

199   287328920 205959480       1395  0.39%  0.42%  0.40%   0 CEF: IPv4 proces

148   1485302401471432555        100  0.07%  0.09%  0.09%   0 ADJ resolve proc

167    31255028  87977151        355  0.07%  0.05%  0.06%   0 DHCPD Receive

192   112436456  52997832       2121  0.07%  0.09%  0.08%   0 IPC LC Message H

I can see which process is causing the jump, I just don't know if this is normal!

Thanks for your time guys!

[amabdelh]

so it is Virtual Exec process, and I beleive you will not see the same issue when you use console/ssh

this is not normal, usually we see similar behavior when you generate the show tech, but show running should not cause high cpu. I beleive you can live with this, it will not cause any impact as this is not a high priority process, but if you like you can upgrade to a different version. good to check the bug list of a particular image before make an upgrade desicision so you will be aware if you are gonna hit any of the knwon bugs.

Regards

[Leo Laohoo]

Multilayer is 3750x, IOS 15.1, L3 is a Catalyst 4000 and the IOS is 12.2.

C'mon now, Edison is asking a legitimate question and your IOS is "15.1" and "12.2"?   Do you have any idea how long the list you've just provided????

Can you be more specific? 

[petenixon]

I had already left the office and had to use my (hazy) memory!!

[Leo Laohoo]

I have a few production switches (x2 multilayer, x1 layer 3) that have minimal CPU usage (at around 20%) but entering a show run will spike the CPU upwards of 90%.

Forget it.  I'll make this nice and sweet.

You running Layer 3?  Use 12.2(55)SE7 or 12.2(55)SE8.  DO NOT, under any circumstances, use 15.0(1) and 15.0(2).

[amabdelh]

Can you try to telnet and do "show runn", do the same thing but with ssh and from console
And please get the output of "show process cpu sorted | e 0.00"

Sent from Cisco Technical Support iPhone App

[jayson.sarausos]

I have this problem 2 years back before with 2960's and 3750's eventually i found out that its just a IOS cosmetic bug.  The high CPU utilization will go upto 99% when you telnet into the machine.