At 8:23 AM -0400 6/14/02, brian kastor wrote:
>All,
>
>I have 2 sites. We currently get full routing tables from uunet and AT&T
>into a single router. We want to put a second router at another site and
>run IBGP between just those two. We will have two routers between the two
>running BGP. I guess it would look something like this:
UUnet AT&T
^ ^
> | |
>site1-r1(bgp)---r2---r3---r4(bgp)-site2
I'd make R2 and R3 route reflectors for R1 and R4 respectively.
Otherwise, you won't have full mesh.
>
>R1 will get UUnet, r4 will get at&t. R1(UUNet) is being used as primary.
>Behind R1 is our HQ. Behind r4 is a large branch.
>My concern would be if the link to UUnet goes down, how does traffic from
>site 1 get out r4.
>
>Scenario:
>User pings 12.x.x.x This is unknown on site 1, so it goes to r1. UUnet is
>down, so the route is pointing to r4 (ibgp). The packet still has a
>destination address of 12.x.x.x, but how does r2 know what to do with this
>packet? It doesn't know where 12.x.x.x is?
Now I'm confused. If you're getting full routes, why should there be
a problem? If the UUnet link goes down, AT&T should know how to get
to UUnet through private peering, etc. In other words, your overall
BGP system has two routes to 12.x.x.x, but the preferred route is
normally through the UUnet link. If that link goes down, as soon as
the BGP system detects it, it will replace that route with a route
via AT&T, which it already knows.
>
>Someone here at work brought this up to me and I keep saying that the ibgp
>session gets established, so when you look at the table on r1 for 12.x.x.x,
>the next hop is the interface of r4. He keeps saying, 'yeah, but when that
>packet gets to r2, the destination is 12.x.x.x, which it doesn't know.'
But R1 will withdraw the route, and all the BGP speakers still have
the route via R4.
>
>All of this can be solved if someone tells me that we just need to run ibgp
>on all 4 routers and i shouldn't see a performance hit. I am worried
>because these routers are going to be routing user traffic between the
sites.
BGP in and of itself isn't a performance hog. BGP performance hits
come more from changes. Yes, a withdrawal of all direct UUnet routes
will cause a brief performance hit.
>
>Thanks,
>
>bk
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=46594&t=46565
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
