High CPU Utilisation

[James Matthews]

Can anyone tell me why I'm getting the following error on one of our 6509 Access switches:

Dec 2 12:46:25.262 UTC: %SYS-3-CPUHOG: Task is running for (2000)msecs, more than (2000)msecs (0/0),process = crypto sw pk proc.

We have around 20 of these switches all configured identically using sw version 12.2-33-SX1I and only one of them is throwing out this error message?

I've tried looking at the page - What Causes %SYS-3-CPUHOG Messages?

This suggests that the issue may be a bug, although when I look at the Cisco bug toolkit, I cannot see any mention of it on there.

Any suggestions would be gratefully received.

[Mahesh Gohil]

Hello James,

CPU hog is caused when some process takes more time interval. It is kind of watchdog to ensure system does not

loked due to total cpu consumption by any process

http://www.cisco.com/en/US/products/hw/iad/ps397/products_tech_note09186a00800a6ac4.shtml

In your case the process crypto sw pk proc.process taking excess of time

It is advisable to log case with cisco

Regards

Mahesh

[Alexander Pai]

CPUHOG errors are simply an informational message that indicate a process has been running longer than the default timer (2000ms).  The idea is to let the user know that a certain process may be taking too long to complete.  It can be an indication of a software bug, but can also be induced simply by load conditions.

In your case, we see that the "crypto sw pk proc" has been busy.  This indicates that we're taking awhile to send out software encrypted ipsec packets.  Encryption is particularly processor intensive without any hardware accelleration.  Therefore, if your crypto utilization is high enough on this device, we could expect to see some degree of CPUHOG errors and poor crypto performance.  Based off of your description, I would guess that it's the crypto traffic rate that is higher on this particular device.

-Alex

[James Matthews]

OK, this does make sense, however, I don't understand why the switch would be using the crypto sw pk process at such regular intervals:

Dec 3 00:46:56.465 UTC: %SYS-3-CPUHOG: Task is running for (2000)msecs, more than (2000)msecs (37/33),process = crypto sw pk proc.

-Traceback= 414E22F8 414E13F0 414E125C 414E1350 414E23D0 414EEC3C 414E2390 414DC458 414EA9DC 414EB98C 414F4B28 414F48E4 414F3B00 414ED72C 414DE680 41

Dec 3 01:46:59.082 UTC: %SYS-3-CPUHOG: Task is running for (2000)msecs, more than (2000)msecs (0/0),process = crypto sw pk proc.

-Traceback= 414E22E4 414E13F0 414E125C 414E1350 414E23D0 414EEC3C 414E2390 414DC458 414EA9DC 414EB98C 414F4B28 414F48E4 414F3B00 414ED72C 414DE680 41

Dec 3 02:47:01.738 UTC: %SYS-3-CPUHOG: Task is running for (2000)msecs, more than (2000)msecs (10/9),process = crypto sw pk proc.

-Traceback= 414E1E90 414E125C 414E1350 414E23D0 414EEC3C 414E2390 414DC458 414EA9DC 414EB98C 414F4B28 414F48E4 414F3B00 414ED72C 414DE680 41409F2C 41

Dec 3 03:47:04.042 UTC: %SYS-3-CPUHOG: Task is running for (2000)msecs, more than (2000)msecs (12/11),process = crypto sw pk proc.

-Traceback= 414E1FF0 414E125C 414E1350 414E23D0 414EEC3C 414E2390 414DC458 414EA9DC 414EB98C 414F4B28 414F48E4 414F3B00 414ED72C 414DE680 41409F2C 41

Dec 3 07:47:16.004 UTC: %SYS-3-CPUHOG: Task is running for (2000)msecs, more than (2000)msecs (65/64),process = crypto sw pk proc.

-Traceback= 414E7A38 414F42B0 414F0964 414ED4EC 414F4B58 414F48E4 414F3B00 414ED72C 414DE680 41409F2C 4140A3A8 41337FE4 41337FD0

As far as I know, the only thing we would be using the crypto process for would be when we ssh to the box. We are not for example using IPSec.

What makes me suspicious more than anything though, is that none of the other switches in our environment are getting this error?