AS-Prepending problem ( Not advertised to any peer)

[Hamed Karimi]

Hey guys,

As you can see R1,R2 and R3 are in AS 100 and have iBGP together also in the same EIGRP. I did AS-Prepneding on R2 for receiving route from R5 (8.8.8.8) on inbound traffic:

But R3 or R1 does not receive it as the redundant or second route from R2:

R3:

R1:

But if I shutdown the iBGP between R2 and R1, then R3 will receive the update:

Or if I remove the next-hop-self it will receive it as well.

Any idea why is like that?

[mlund]

Hi

That's because R2 only advertise the best path. When you disable link R1-R2, R2 would only have one path left, and that will be the as-prepend path, and then it will be that one advertised.

/Mikael

[Hamed Karimi]

But when we have the iBGP between R1-R2,

in R2 we can see it has 2 available path.

but R2 even dose not advertise the best path (R3->R2->R1-R5) to R3.
If R3-R1 link goes down, R3 can not reach 8.8.8.8 via R2 since it did not get any update from R2.

[Jon Marshall]

Just to add, the usual way to influence outbound traffic is with local preference. 

You can use AS path prepending if you want but that is more commonly used to influence traffic into your AS. 

Jon

[Hamed Karimi]

What I just realized is that, R2 is not advertising the best path (via R1) to R3, because it gets it from an ibgp. So it can not give that update back to R1 itself or another iBGP which is R3.

correct?

So what is the solution? If the ISP put prepending, I wanted to have full mesh iBGP, but in this case if R3-R1 link goes down, the traffic will not go to R2.

[Jon Marshall]

Correct in what you say. 

You can make R2 a route reflector but even that will not work as is because the next hop IP for the route from R1 will be R1's 172.20.20.x IP and R3 will not have a route for this so it would never use it. 

What exactly are you trying to do in terms of traffic paths ie. do you want R3 -> R1 -> R5 to be the active path whilst everything is up and running ? 

Is it just R3 internally or do you have other routers etc. 

Jon

[Hamed Karimi]

I wanted to have R5-R1 as primary for internal network (R3).

but if R3-R1 link goes down, R3 can not send to R2. And if I remove the ibgp between R3-R2 even if R3 receives the traffic, it will send it directly to R5 (not preferred)

I assume I should not extend the BGP to R3. And using IGP redistribution on R1 and R2.

or if I extend it like the current topology, I should remove the R1-R2 link. However we will not have the R5-R1 as primary path for internal network.

[Jon Marshall · Accepted solution]

As usual there are a number of ways to achieve what you want - 

1) advertise the 172.20.20.0/24 subnet into BGP (but make sure not to advertise outside of AS) and then make R2 a route reflector 

2) use an IGP and not run IBGP on R3 but you would still need to modify metrics so R3 to R1 is preferred 

3) if it is only R3 and no other internal routers you could simply remove the AS prepending on R2, let R3 receive equal cost paths from both R1 and R2 and use weight to choose the R1 route. 

and probably others I have not thought about. 

It really depends on the rest of your network and exactly what you want to achieve and whether you see more internal routers being used later etc. 

One thing I would do though is if you want to modify the BGP attributes for the incoming routes use local preference and not AS prepending as you currently have done. 

Jon