Giuseppe Larosa Wed, 09/30/2009 - 05:37

Hello Oleg,

there are some global commands like

scheduler ?

allocate Guarantee CPU time for processes

heapcheck Extra checking to validate memory

max-sched-time Maximum time scheduler can run without flagging an error

process-watchdog Action for looping processes

runtime Process runtime controls

switch Allocate CPU time on SP

usually you can provide a max time of usage for an interrupt with.

first option may be interesting for you

scheduler allocate ?

<400-60000> Microseconds handling network interrupts

scheduler allocate 10000 ?

<100-4000> Microseconds running processes

But before trying a tuning like this you should try to understand if there is something that can be done at the configuration level to reduce cpu load.

I mean if you have multiple BGP full tables or at least one for a few seconds cpu goes to 100% when checking BGP next hops for example.

if you post

sh proc cpu sorted 1min

sh ip bgp sum

it is possible to have an idea if what is happening is normal or not.

Hope to help

Giuseppe

181270aladdin Fri, 10/02/2009 - 05:58

Hi Giuseppe!

Thank you very much for your ansver!

Unfortunately, I couldn't found manual about it for beginner. The official manual is very compressed and ,essentially, havn't difference from

"scheduler allocate ?" :-)

I found this example in the manaul

"This example shows how to make 20 percent of the CPU available for process tasks:

Router-config# scheduler allocate 2000 500"

But I can't understend, how are they made percents from seconds!?

PS: I wrote about BGP for example.

"sh ip route | include x.x.x.x" made too if the router have full BGP. I have loading CPU 100% for 20-40 sec.

Joseph W. Doherty Fri, 10/02/2009 - 15:26

100% CPU isn't always a bad thing. Do you believe these peaks are causing some performance issue or are you just concerned that they are happening?

181270aladdin Sun, 10/04/2009 - 21:03

Hi Joseph!

Of course, I know that Cisco allow of brief 100% CPU. For example, when my uplink bgp peer go down, the router have 99-100% CPU and it's normal (may be?).

But if I reload the router, I have 4 full BGP from uplinks and 30 BGP peers to my downlinks. Consequently, the router cycling reloading, and I think it's abnormal :-)

Joseph W. Doherty Sat, 10/10/2009 - 08:04

There may be a CPU spike when BGP peering is being re-established. If the BGP tables are large, e.g. Internet routes, might also expect the BGP scanner to spike the CPU when it runs. The latter, if I recall correctly, runs once a minute.

Again, what you might be seeing might be normal and not cause for concern.

Actions

This Discussion