Router Doesnot pass e-BGP Routes to IBGP neighbor

dinesh.thathanath · ‎03-23-2010

Hi Folks,

I am facing a very interesting scenario on my routers.

I have Routers R1 and R2 connected as mentioned below :

R1 -> MPLS SP1 - E-BGP Peering

R2 -> MPLS SP2 - E-BGP Peering

R1 -> R2 - Back to Back Connection for IBGP

On R1 :

I am learning a Route 10.0.0.0/11 from MPLS SP1. This route is tagged with a local prefence value of 200 and passed on to R2.

On R2 :

I am learning 10.0.0.0/11 from MPLS SP2 (This route is prepended and advertised from a different location). Interestingly this route is not advertised to my IBGP peer.

I am attaching a PPT and the configs for better understanding. Any idea on this guys?

Regards,

Dinesh

milan.kulik · ‎03-23-2010

Hi,

a BGP router is always advertising only the best route to a particular prefix to his neihgbors.

On R2, the best route to 10.0.0.0/11 is that received from R1 (local preference 200), as you see below:

R2# sh ip bgp
BGP table version is 2820, local router ID is 172.30.22.253
Status codes: s suppressed, d damped, h history, * valid, > best, i - internal,
r RIB-failure, S Stale
Origin codes: i - IGP, e - EGP, ? - incomplete

   Network          Next Hop            Metric LocPrf Weight Path
*>i10.0.0.0/11      172.22.129.2             0    200      0 9829 65001 i
*                   10.1.70.25                             0 4755 65002 65002 65002 65002 65002 i

So R2 will not advertise the route received from the other eBGP neighbor 10.1.70.25, as it's not the best one.

HTH,

Milan

dinesh.thathanath · ‎03-23-2010

Thanks Milan.

Is this some sort of a split horizon implementation on BGP?

Since the R2 is not updating about the routes to R1, does it increase the convergence time (since R1 BGP table doesnt have the entry)?

Is there any command which will show me that the routes are not advertised to R1 and due to this reason?

Thanks,

Dinesh

milan.kulik · ‎03-23-2010

Hi,

I would not call this a split-horizon, it's a basic BGP feature: Only best routes are advertised to the neighbors.

(If all the BGP tables would be advertise, just imagine how much memory would be required for the whole Internet BGP table.)

Regarding convergence time - you can decrease it by tuning BGP timers.

The default timer values are designed to make the Internet as stable as possible, if your private network includes only a small number of prefixes you can decrease the timers. But it's an advanced BGP excercise and might require your provider cooperation.

If you issue "sh ip bgp" command, only the lines starting with > character are advertised to the neighbors.

I hope it answers your third question.

HTH,

Milan

dinesh.thathanath · ‎03-23-2010

Just out of curiousity :

Can this behaviour be changed by issuing any BGP commands (maximum paths or something like it?).

My primary objective was to check on R1 to see whether I have a backup route available through R2. Seems like it is not possible in my scenario I guess..

Thanks,

Dinesh

milan.kulik · ‎03-23-2010

Hi,

I don't see any way if you want to keep R2 using R1 as next-hop to 10.0.0.0/11.

You could use weight instead of local preference on R1 and make the eBGP neighbor the next-hop on R2 configuring his weight.

Then R2 would advertise his (eBGP) best route to R1.

You would still keep R1 using his eBGP path to 10.0.0.0/11 but you would see the backup route via R2.

But any traffic sent to 10.0.0.0/11 through R2 would be sent through AS 4755.

BR,

Milan