The MPLS-OPS Archive

[Rajiv Asati <rajiva@cisco.com>]

Amos,

At 07:22 AM 7/22/2002, robert wrote:
> Hi,
> I'm building an mpls vpn in the following
> topology:
>  
> Ce1-Dsl  pppoA connection - Pe1- wan
> connection-Pe3 - F.R connection - Ce3
>                                              
> |                              
> |
>                                              
> |                              
> |
> Ce2 - Dsl pppoA connection - Pe2 - wan
> connection--- -
>  
>  
> Pe1 and Pe2 are also connected between them but
> there is no mpls runing there.
> mp-bgp is runing between pe1 and pe3 and
> between pe2 and pe3 but not between pe1 and pe2. pe3 is a route reflector
> for pe1 and pe2.
> Pe1 knows ce2 addresses through mp-bgp and pe2
> knows ce1 addresses.
> I do have connectivity between Ce1/pe1 to ce3
> and also from ce2/pe2 to ce3, but not between  ce1 to
> ce2.
> i think that the reason is that in pe1, ce2
> routes are seen with next hop of pe2, and as i said there is no mpls vpn
> between them so no vpn conectivity.
> what i tried to do is to issue the following
> command in pe3: nei pe1 next hop self. pe3 accepts this command, but in
> pe1 the routes are still seen with pe2 ip
> address.

That's the correct behavior. 
RouteReflector should not change the next-hop of iBGP learned route,
however it would do so if it was an eBGP learned route.

> Is there any
> restrictions to using that command in mpls enviorment
> ?

I don't think that MPLS has anything to with such behavior. :)
We should see the same for IPv4 route as well.

>   
> Any help/idea will be
> welcomed.

Well, simple thing could be to either shut down the link between PE1 and
PE2, or turn MPLS on it.
Or you may want to tweak the IGP cost associated with PE1 (from PE2's
perspective) and make PE1 reachable via PE3. Do the same for PE2 (from
PE1's perspective).

Rajiv

>  
>  
> Thanks
> Amos
>  
>