BGP loop

[Amos Kafwembe]

hello all,

i have a problem with my BGP.

my network topology is as follows; my LAN<------------->router1(BGP)<----------->Multilayer Switch (with IP routing)<------------------->router2(BGP) - IXP

my two routers are configured as eBGP peers and the BGP is even running. the problem is; from my LAN, when i trace a local websites IP address learned through BGP from the IXP router i get a loop between the MLS and Router1. i have a default route on the MLS pointing to Router1's IP add, so am guessing the MLS doesnt know how to get to the routes learned by BGP through the IXP router and keeps sending the data back to through the default route. is there a way i can tell the BGP to ignore the MLS??? i have tried with next-hop-self but to no avail.

kindly assist.

regards.

[mvsheik123]

Hi Amos,

Your finding sounds correct with ref to 'loop' issue. R2 is sending to R1 a default route? Is there nay way you can change the default route on MLS to R2?

Thx

MS

[Amos Kafwembe]

hi mvsheik123,

router 2 is directly connected to the MLS and there is no route pointing to it *since its directly connected* the peering between R1 and R2 is working, but as i said i cannot trace IP addresses or networks learned through BGP, when i try to trace i am getting a loop between MLS and Router1. its like the MLS is not passing the traffic on to the IXP router.

[Alessio Andreoli]

Hi Amos,

It is not really clear to me the topology of your network. Do you have two AS in the same company?

By the way the ping pong you see with the trace echo are very likely a problem with routing and not a loop. This happens when for example you have a static route showing to the packet the outbound direction while no route is pointing the inward direction. When a router does not know a direction it keeps asking until the TTL expires. As per your case it looks something wrong with redistribution or with BGP routing. If you posted more info it would be easier to figure it out.

I would start to check the routing between mls and your inner router.

Hope this helps

Alessio

Sent from Cisco Technical Support iPad App

[Amos Kafwembe]

hi Alessio Andreoli,

yes we have 2 AS...at the IXP we have a different on from at Router 1 hence we have eBGP peering between the two. on MLS i have a default route pointing to router1 and i think this is the reason i get a loop. the MLS doesnt know the routes learned through BGP by router 1 and hence when router1 forwards packets to those destinations through the MLS, the MLS sends them back to router1 using the default route, since it doesnt know about them.

is there a way i can bypass the MLS so that routes learned through BGP can use router2 interface as the next hop?

[cadet alain]

Hi,

correct me if I'm wrong but the MLS switch doesn't participate in BGP and you have an eBGP session using Loopbacks between R and R2 ? In this case the next-hop is already R2 bu to get to R2 , R1 does a recursive lookup and sees it must send packets to MLS, now MLS when it sends packets to the remote prefixes advertised by R2 does a routing lookup and it has only a default route going to R1 and so it sends back packets to R1 and so on so there is a loop.

I think that one solution could be making R1 and R2 peer through a GRE tunnel which would take out the switch from the equation.

Regards.

Alain

Don't forget to rate helpful posts.

[Amos Kafwembe]

Hi cadet alain

yo analysis is exactly as you have put it! and your advised solution is what i think may work. now my question is, how do i peer through a GRE tunnel? can u advise with a sample config?

[cadet alain]

Hi,

just do a GRE tunnel between the 2 routers and then peer on the IP addresses of the tunnel endpoints.

shutdown your eBGP peerings you already have with the neigh x.x.x.x shutdown command then

eg:

r1

int tu0

ip add 10.1.1.1 255.0.0.0

tuneel source x/x

tunnel destination x.x.x.x

keepalive

exit

router bgp xxx

neigh 10.1.1.2 remote-as xxx

r2

int tu0

ip add 10.1.1.2 255.0.0.0

tuneel source x/x

tunnel destination x.x.x.x

keepalive

exit

router bgp xxx

neigh 10.1.1.1 remote-as xxx

Regards.

Alain

Don't forget to rate helpful posts.

[Amos Kafwembe]

heya.

thanks for that, I will be sure to try this out when we have a window of maintenance open for us. will update and rate!!! thanks!

[milan.kulik]

Hi,

another possibilities would be:

b) let the MLS switch participate in BGP

c) redistribute BGP prefixes from your routers to an IGP (OSPF, e.g.) running on the MLS switch.

HTH,

Milan

[Alessio Andreoli]

Hi Amos,

i think that all what you have to do is to change the routing config on the MLS. If you posted its config i could help you more. Essentially, you should be able to establish layer 3 connectivity between the Router 1 and the Router 2 loopbacks. After that you should state below the BGP process

neighbor 1.1.1.1 source loopback 0 !(or whatever number you have chosen)

and that should be what you need. The issue, it is my understanding , is on the MLS routing decisions. Before implementing any other technology ( a good suggestion is from Alain) make sure network reachability is in place otherwise nothing will work.

Take Care

Alessio

[Amos Kafwembe]

Hi Alessio,

basically the routing on the MLS is all static routes, pointing to various networks and a default route to the router 1. Network reachability and connectivity is there. right now i have normalised the network, and the switch is back to layer 2 forwarding only.