BGP Peering in PrivateIP MPLS - CPE-to-PE only? Or also CPE-to-CPE?

Answered Question
May 12th, 2008

Maybe a dumb question, but I'm not very experienced with BGP (yet).

We are implementing multi-site Private IP MPLS. We are peering to the providers equipment using BGP.

Do we also need to configure all our other routers as BGP neighbors of each other? Or does each only need to have the provider's router configured as a neighbor?

The reason this is coming up, we first configured only the Provider as the BGP neighbor, but we didn't see our internal routers propagate. When setting up our own routers as neighbors of each other we now see our routes, but in some cases we are introducing routing loops between the PE and CPE routers.

I have this problem too.
0 votes
Correct Answer by Jon Marshall about 8 years 7 months ago

Eric

You should not need to have your CE's peering with your other CE's. Indeed that is one of the advantages of MPLS in that you do not need all your CE routers maintain peerings with all your other CE routers.

All you should need is a peering between your CE and the Telco PE device. The PE devices then exchange your routes and propogate them to your other CE devices.

Have you checked the basics ie.

On a CE device check that

1) You are advertising out your internal routes to the Telco PE

sh ip bgp neigh advertised-routes

2) You are receiving routes from the PE

sh ip bgp neigh received-routes

OR

"sh ip bgp" will show you the routes in the BGP routing table.

3) If the networks from a remote site are in the bgp table check your redisribution into OSPF.

Jon

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Loading.
ericjwhite Mon, 05/12/2008 - 19:14

Hi M,

The configs are pretty straghtforward and likely wouldn't help much - just BGP neighbor statements.

More detail is easy however. Multiple sites, each site looks like this:

L3switch---CPE-RTR---PE-RTR---mplsCloud

OSPF between L3Swtich and CPE-RTR, BGP between CPE-RTR and PE-RRT, redisribuution from OSPF to BGP and visa versa.

Now picture that in each site, so you have OSPF internally, being redistributed to BGP on all my CPE routers. BGP from my router into the PIP core, and also BGP on my router being redistributed into OSPF on the ingress.

So we have IP connectivity between CPE routers via the IP scheme provided by Telco. Those same routers have the ability to peer to one another over BGP, which is how we got our internal routes to propagate in the first place, but should we be letting Telco PE routers handle that part of the routing? or should we create peers between all our CPE routers? By creating peers between our CPE routers we are going through the MPLS network to do so, like this:

CPEpeer1--PEpeer1--MPLS--PEpeer2--CPEpeer2

Where CPEppeer1 is neighbors with PEpper1 AND CPEpeer2, and visa versa.

Multiply this by several sites, and you have mutlple routers routers peering to both the provider cloud AND each other (through the provider cloud).

Jeeze this sounds more complicated than it is - hope it makes sense!

Correct Answer
Jon Marshall Mon, 05/12/2008 - 22:22

Eric

You should not need to have your CE's peering with your other CE's. Indeed that is one of the advantages of MPLS in that you do not need all your CE routers maintain peerings with all your other CE routers.

All you should need is a peering between your CE and the Telco PE device. The PE devices then exchange your routes and propogate them to your other CE devices.

Have you checked the basics ie.

On a CE device check that

1) You are advertising out your internal routes to the Telco PE

sh ip bgp neigh advertised-routes

2) You are receiving routes from the PE

sh ip bgp neigh received-routes

OR

"sh ip bgp" will show you the routes in the BGP routing table.

3) If the networks from a remote site are in the bgp table check your redisribution into OSPF.

Jon

ericjwhite Tue, 05/13/2008 - 19:03

Jon, thank you for the detailed reply! Very much appreciated, I believe this solves my issues.

ericjwhite Fri, 05/16/2008 - 17:22

Jon, sorry for the late reply. Thank you for the detailed info! Now that I've sorted out my route redistribution everything is working great with just the peer between the PE and CPE.

Thanks very much for the help!

Actions

This Discussion