[c-nsp] MPLS Label propagation and/or MP-BGP and/orroute-reflecting, oh my.

Pshem Kowalczyk pshem.k at gmail.com
Fri Jan 21 18:18:24 EST 2011


Hi,

On 22 January 2011 09:50, Jason Lixfeld <jason at lixfeld.ca> wrote:

{cut}

>> Your PE1 and CE are in the same AS, and IOS does not support iBGP as
>> PE-CE routing protocol.. So this is a one problem. The issue with lack
>> of iBGP PE-CE support is the fact that the vpnv4 next-hop is not
>> rewritten to the PE1's loopback, so the other PE's in the network will
>> not find an LSP towards the unchanged nexthop, but you can somehow make
>> this work re-writing the vpnv4 nexthop before sending the vpnv4 update
>> to the peers. Still unsupported, though.
>
> Any method that would be more supported?
>

We ran into the same issue, with our non-MPLS network being migrated
to MPLS .Even though officially not supported we use iBGP as PE-CE
protocol. Both routers have 'route-reflector-client' enabled and
'next-hop-self'. If you have multiple PE-CE links you have to ensure
that you control the  re-advertisement of routes received from one
domain through the other and then back to original domain
(non-MPLS->MPLS->non-MPLS and the other way around).
We've tested that with PE being 6500 (can't remember the exact
software version, one of the SXH ones) and ASR9k (3.9.1).
Both CsC and standard L3VPNs are good if you don't have to run the
same services on both networks, effectively when the new MPLS core is
a completely separate cloud. Also, in some circumstances OSPF might be
an option (but not for internet routes).

I agree that this is not 'proper' design, but sometimes there is no
other choice.

kind regards
Pshem


More information about the cisco-nsp mailing list