Re: BGP Keepalive Timers

we have a similar extranet but with more eBGP peers around 100 and we use
timers 1 3. Peers are usually ISRs and ISR2 (so no bfd option) and we get
around 10 to 15 prefixes per peer. Hw is 7609 with RSP720 and we are stable
even with 20 peers flapping simultanously.
 But when the aggregating interface flaps and cause all peers to flap also,
cpu goes into a loop and doesnt come back hence we shut down all peers and
bring them 10 by 10 with 30seconds intervals.

My question is, can we make EEM script for such a case to see the bgp peer
flappings and shutdowns peerings and after a custom time it goes and say no
neighbour x.x.x.x shutdown? Any suggestions? By the way each peer has its
own vrf.

Thank you!

Serdar Kut
CCIE2#18130 (RS,SP)

On Thu, Apr 17, 2014 at 4:19 PM, Brian McGahan <bmcgahan_at_ine.com> wrote:

> It's because the BGP process and its keepalives are software switched on
> the CPU. If the CPU gets busy with something like BGP convergence then the
> session will flap. BFD offloads the processing in hardware to the linecard,
> so you get fast linkdown detection without the danger of the CPU
> accidentally dropping the session.
>
> This is true of all protocols, e.g. OSPF, EIGRP, IS-IS, etc. It's not a
> BGP problem it's a process switching problem.
>
> Brian McGahan, 4 x CCIE #8593 (R&S/SP/SC/DC), CCDE #2013::13
> bmcgahan_at_INE.com<mailto:bmcgahan_at_INE.com>
>
> Internetwork Expert, Inc.
> http://www.INE.com<http://www.ine.com/>
>
> On Apr 17, 2014, at 4:50 AM, "Dave Serra" <maybeedave_at_yahoo.com<mailto:
> maybeedave_at_yahoo.com>> wrote:
>
> Hi Guys,
>
> We have an extranet with about 20 or so partners and we peer BGP
> with them. Combined we only have a few hundred routes in the BGP table
> from
> them. While in conversation with one of them for new connectivity I
> wanted to
> use keepalive timers of 1 and 3 and the provider said this is way too small
> and will produce instability in BGP meaning BGP may flap. Sure enough
> when I
> tried to configure it I also got a warning that a hold down timer of less
> then
> 20 seconds would cause BGP to flap. My question is why? I keep hearing
> the
> answer that the circuits could flap and trigger failover but failover is
> desirable as the application has a 30 second timeout before it resets the
> TCP
> connection. I told this to the provider and he was talking about the BGP
> process not being able to process all of the keepalives but using BFD we
> could
> do this. I don't get why the BGP process would have an issue with 20 or 40
> BGP sessions all sending 1 keepalive per second.
> This is such a small amount of traffic. Someone at work also mentioned
> there
> may be policing on the cat6500 that keeps all of the hellos from getting to
> the CPU and too many sessions can cause ANY of your BGP sessions to flap
> but I
> that also doesn't sound right.
>
> Anyway, if anyone knows why the BGP process
> wouldn't be able to handle 1 and 3 keepalives for say 20 or even 40 peers I
> would be very interested to hear more.
>
> Thanks again!
>
>
> Make a small loan,
> Make a big difference - Kiva.org<http://Kiva.org>
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Thu Apr 17 2014 - 19:34:51 ART