[nsp] intermitten ping lags on 7500/rsp4/256M

[John Murphy]

Once every 60 seconds BGP Scanner walks the entire RIB.  Part of what it
does is validate next-hop reachability for each entry.  This is a well
documented source of high CPU on routers.  Take a look at
http://www.cisco.com/warp/public/459/highcpu-bgp.html#topic2 it speaks
specifically to ICMP and apparently high latency regarding pings directed to
the router.

It's not so much a matter of difficulty as it is a matter of priority.
Would you rather (a)continue passing traffic or (b)respond to a ping.

HTH

John

----- Original Message ----- 
From: "Jared Mauch" <jared at puck.nether.net>
To: "Dmitri Kalintsev" <dek at hades.uz>; <cisco-nsp at puck.nether.net>
Sent: Thursday, June 12, 2003 6:48 PM
Subject: Re: [nsp] intermitten ping lags on 7500/rsp4/256M


> On Fri, Jun 13, 2003 at 08:58:57AM +1000, Dmitri Kalintsev wrote:
> > You can't "fix" it. ICMP handler is low-priority process in IOS. These
> > "delays" are only evident when you try to get ICMP replies from the
router
> > itself, but they do not affect the transit traffic.
> >
> > On Thu, Jun 12, 2003 at 04:41:23PM -0400, lexus wrote:
> > > Chris,
> > >
> > > How did you fix yours?
> > >
> > > What IOS image are you running?  I am willing to try anything =)
> > >
> > > Thanks
> > > -ken
> > >
> > > -----Original Message-----
> > > From: cisco-nsp-bounces at puck.nether.net
> > > [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Cisco Geek
> > > Rotation
> > > Sent: Thursday, June 12, 2003 3:43 PM
> > > To: cisco-nsp at puck.nether.net
> > > Subject: RE: [nsp] intermitten ping lags on 7500/rsp4/256M
> > >
> > > At 02:15 PM 6/12/2003 -0400, you wrote:
> > > >Ive seen the same thing.  If you get on the 7500 and do a..
> > > >
> > > >Router# sho proc cpu | excl 0.00%
> > > >
> > > >You will probably see BGP updates jumping in when you hit that 600ms
> > > >ping.  At least that is my experience.  How to fix it (if you
consider
> > > >this broken) is a very good question :)  I'm open to answers.
> > > >
> > > >-Scotty
> > >
> > >
> > > Ditto here, it's caused by BGP on my 7507.
>
> What you need is the distributed linecards to inspect packets
> prior to being forwarded to the RE and be able to generate the icmp
> responses at linerate back.  Obviously non-distributed
> platforms (3640, 2600, 2500, 7200, or even 7500 w/ older non-dcef
> capable cards) will not be able to respond in this fashion, but they
> are more likely to not be performing bgp.
>
> This doesn't seem like a too complicated thing to do.
>
> - jared
>
> -- 
> Jared Mauch  | pgp key available via finger from jared at puck.nether.net
> clue++;      | http://puck.nether.net/~jared/  My statements are only
mine.
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> http://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/