From: CCIE3000 (ccie3000@googlemail.com)
Date: Mon Aug 11 2008 - 14:32:28 ART
Hi Joe,
Sorry for the delay in getting back, I've been bashing this one to bits.
Originally I was expecting R5 to be the best prefix, so then I started
really looking at the route decision process.If you don't change the
next-hop on R3 then R6 has R5 as the best path.
I'm assuming that because it has two routes both with the same next-hop
address and it knows that the next hop address is on the same broadcast
domain it doesn't even bother looking into using R3. If anyone can put this
in a more technical way please shout out :)
Next I changed the next-hop on R3 and because it it has the lower router ID
R6 chose R3. When I gave R3 a higher router id, R5 became the router
of choice. I also connected R3 and R6 on a separate broadcast domain, same
again, which-ever had the lower router ID was R6's choice.
*update: whilst writting this I've been playing some more..*
**
Basically you can leave the next hop unchanged and give R3 the higher weight
(or local preference etc etc ). In the BGP table of R6, R3 will be numero
uno, when you do a show ip bgp on R6 it will show the route from R3 as best
but the next hop address will still be R5, when you do a show ip route *
bgp prefix* it will show it via R5, obviously although the bgp table show's
the R3 router to be the best, when you do a trace to the bgp address it goes
directly to R5.
All the above is simple, silly stuff but sometimes it's easy to overlook the
simple things, hopefully the above gives us a bit of food for thought to
others.
Cheers,
Si.
On 8/11/08, Joseph Brunner <joe@affirmedsystems.com> wrote:
>
> R6 See's the routes from both R3 and R5 but it sees all next hop-addresses
> as R5.
>
> Why is this?
>
> >R6 is choosing R5 as the *best* route for the routes originally learned by
> >R3/R5 from R1, BB1
>
> You are seeing the correct behavior.
>
> All the paths should have two entries
> Such as
>
> Show ip bgp
>
> 1.1.1.0, 3.3.3.3
> >1.1.1.0, 5.5.5.5
>
> But the ">" is on the routes to R5, yes?
>
>
>
> When R3 does a recursive look up for
> the next-hop-address it sees it via R5 and hence this is why it's
> advertising the bgp routes to R6 with the next-hop address of R5?
>
> >not exactly. EBGP peers simply REset the next hop to the update source by
> default, IMHO. However since you mentioned this is on a broadcast segment,
> its possible.
>
> Can you try to make R6 only directly adjacent to R3 (say via a p2p link in
> Dynamips), remove the R6 -> R5 peering and see what the next-hop's look
> like...
>
> Deep down I remember something like this YEARS back on a /24 frame-relay
> cloud!
>
> -Joe
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
> CCIE3000
> Sent: Monday, August 11, 2008 5:29 AM
> To: Cisco certification
> Subject: Silly BGP next hop question
>
> Hi Guys,
>
> I've got a silly question concerning next-hop with bgp
>
> R1 Peers with BB1
> R1 is the route-reflector for R3 and R5
> They are peered via their loopbacks.
> They see the bgp routes with the next-hop-address of BB1.
>
> R3 and R5 then have an EBGP peering with R6.
>
> R6 See's the routes from both R3 and R5 but it sees all next hop-addresses
> as R5.
>
> Why is this? okay, it's an EBGP peering so the next-hop value should be the
> ebgp peers.
>
> R3 must go via R5 to get to the BGP routes learnt from the BB router.
> R3, R5 and R6 are all on the same subnet.
>
> Is it because in it's BGP table R3 sees the bgp routes with the next-hop of
> the BB router (because it's learnt from the Route Reflector which is ibgp
> peering so the next-hop won't change). When R3 does a recursive look up for
> the next-hop-address it sees it via R5 and hence this is why it's
> advertising the bgp routes to R6 with the next-hop address of R5?
>
> I hope the above is clear enough and not too confusing.
>
> Thanks Si.
>
>
> 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
This archive was generated by hypermail 2.1.4 : Mon Sep 01 2008 - 08:15:30 ART