iBGP - RR

[tckoon]

Based on attached diagram, R1 peer with RR1 via directly connect interface, R2 peer with RR2 via directly connected interface.

If the interface, eg R1 to RR1 interface is down, then R1 will no longer any BGP routes.

What if I iBGP peer R1 and R2 via their back to back interface ? So when interface between R1 and RR1 is down, R1 can still learned BGP route from R2.

My concerned is iBGP peering between R1 and R2 caused routing loop?

Regards

[289114]

Hi,

Even if you peer R1-R2 and the link between R1-RR1 goes down, R1 will not be able to get any BGP updates coming from the Internet. Reason because R2 is learning updates from IBGP peer, will not propagate these updates to another IBGP peer ( IBGP  split horizon rule).
Either specify R1 as a route-reflector and RR1 as its route reflector client
or specify R2 as a route-reflector and RR2 as its route reflector client.

I assumed before that in case the internet link of RR1 or RR2 goes down, you don't need routes to be propagated from the IBGP to the routers RR1 or RR2.

For routing loops, I think the easiest is to use on R1 weight to increase the weight of neighbor RR1 and on R2 weight to increase the weight of RR2.

Regards
Sayed

[289114]

Hi Again,

Sorry , I think you need to increase the weight on the routers RR1 and RR2. I have to verify it.

Take care

[Kishore Chennupati]

Hi,

You dont need to use any RR's here. If you are having a full mesh then the RR1 will advertise the eBGP routes to the rest of routers via IBGP.

Similalry, RR2 will advertise the e-BGP routes to the rest of the routers via iBGP. Remember, for the BGP peering, the routers needn't be directly connected.

Now, regarding which exit router will be used, you need to decide if you want RR1 or RR2 as the primary entry or exit point. and the other link becomes secondary. In order to do this, you just need to increase the local-preference of the router to say 200 which you want primary and advertise the local routes with an AS prepend on the other router for the AS path.  The default local-pref is 100 so all the other routers will choose the one with a high local-pref.

Based on attached diagram, R1 peer with RR1 via directly connect interface, R2 peer with RR2 via directly connected interface.
If the interface, eg R1 to RR1 interface is down, then R1 will no longer any BGP routes.

Here R1 will learn the routes from RR2 via iBGP not R2.  I have attached your diagram with some changes.

What if I iBGP peer R1 and R2 via their back to back interface ? So when interface between R1 and RR1 is down, R1 can still learned BGP route from R2.
My concerned is iBGP peering between R1 and R2 caused routing loop?

There wont be any routing loops as R1 will not learn the BGP routes from R2.

Please let me know if you have any further questions

HTH

Regards

Kishore

Please rate if helpful

[tckoon]

Hi Kishore,

Thanks for the explanation. But if I do not want to have cross-connect between routers(due to the availability of physical link)  then the only way is running IGP and advertised loopback IP then peer using loopback IP. But I have the contraint in my design is I do not want to run IGP between R1/R2 and RR1/RR2.

My contraint are:

1) The physical link as shown as my diagram, no additional link.

2) No IGP between RR1/RR2 and R1/R2.

Do have any workaround to my limitation ?

Regards

[Kishore Chennupati]

Hi,

You dont cross connect them physically. The link is logical. The traffic still physically flows through RR2 to R2 to R1. I am talking about the iBGP peering. You can peer RR2 and R1 without directly connecting them. The link that I put in the diagram is logical not physical.

IGP is used for BGP next-hop reachability.

you can use the directly connected physical links to peer with the routers but what if you change the ip addressing in future. In a iBGP mesh, loopbacks are preferred way to peer as you don't often change them.

HTH,

Regards,

Kishore

Please rate if helpful

[289114]

Hi,

sorry for interference,

If you like to apply the route reflector, it is one command on router R1: (I am not sure if your router R1 or R2 are route-reflector capable) but try this if you want on the router R1

router bgp xxx

neighbor IP_ADDRESS_RR1 route-reflector-client

As your diagram shows, configuring only one router either R1, or R2 as a route-reflector is enough to solve the IGP propagation in case of link down.

Regards,

Sayed