I'm considering a limited IPv6 rollout using 6PE, and would like to know if the following configuration would be supported.
Our typical topology consists of customer aggregation routers connected to a pair of regional core routers. The cores are directly interconnected to other regions' cores via our long haul transport network. Our border routers are also connected to the closest pair of cores.
All the regional pairs of core routers are route reflectors, and the customer aggregation and border routers are route reflector clients of the closest pair of cores.
These are all 7600-class routers, and they're all running MPLS on the interconnects.
Can we use 6PE to exchange V6 routes between our dual-stack customer aggregation and border routers, and have those routes reflected through MPBGP sessions to the IPv4-only core routers?
customer_agg (dual v4/v6) ---> region1 core (v4 only) --->
<--- region2 core (v4 only) ---> border (dual v4/v6)
At a minimum, you need to enable "ipv6 unicast-routing" before you configure BGP to act as a RR for ipv6. IPv6 doesn't need to be enabled on any interfaces though.
Stafford, yes you can run 6 PE in you distributed RR topology. Only thing is even though your RR's need not be dual stack but they definately need to support V6 Address Family for peering with the border PE's and reflecting the updates received.
Do note that when you enable ipv6 address family it makes it a dual stack router even though you do not aggregate customers over ipv6 link addresses.(7600'shave support for this, so should not be much of a concern here in terms of support)
Here is a reference for 6PE configuration.
Other option which you can think of is rather than creaing a full mesh of v6 peerings, if your dployment is limited to certain PE's you may want to form direct v6 ibgp between them.