[rbak-nsp] mBGP / non 6PE / mapped v4 address next-hop

Olivier Benghozi olivier.benghozi at wifirst.fr
Fri Jun 22 12:14:16 EDT 2012


Hi,

> I DONT want to do 6PE - so why would i want to create labels?

It seemed more probable to me that you wanted 6PE, but you really don't :)


>>> The neighbor ASR9k sees the route as:
>> 
>> On the Cisco IOSXR, you must use "address-family ipv6 labeled-unicast" for
>> the neighbor/peergroup, and "address-family ipv6 unicast" with
>> "allocate-label all" under in the BGP router global level.
> 
> I DONT want 6PE - Our core is dual stacked so i'd do everything native. 

As told in http://tools.ietf.org/html/rfc2545#section-4 , BGP next-hop has something to do with the BGP transport, it's not independent, and it depends of the BGP implementation.
I didn't try this; but from what you copy/pasted, it looks like SmartEdge implementation behaves like JunOS with the use of derived "IPv4-compatible IPv6 prefix".

A solution would be to manually fix the IPv6 nexthop by using a route-map for outbound IPv6 routes on your session.
Another solution is to create additional IPv6 addresses on your interfaces, using those "IPv4-compatible IPv6 prefix".

Of course, a solution is to add a second set of iBGP sessions using IPv6 transport.
And finally there is 6PE.

If you want to stay purely native and avoid manual tricks, a second set of IPv6 BGP sessions may be the proper way to go...



regards,
Olivier Benghozi
Wifirst




More information about the redback-nsp mailing list