MP-BGP does not have a requirement to use IPv4 for transport when exchanging IPv6 routing information. IPv6 can be used for transport, too.
Observe the following configuration:
Consider the previous example, where an IPv4 session was used to exchange both IPv4 and IPv6 routing information. The difference here may be that we create two separate TCP sessions, where here, we are configured for an IPv6 transport with IPv6 NLRI. The drawback is increased resource utilization of the router, as it is much more efficient to use MP-BGP to transport multiple protocols over the same transport.
Note the neighbor must still be activated within the IPv6 address-family.
By default, BGP sets its router ID to the highest IPv4 address of any loopback interface, and if none exists, the highest IPv4 address of any physical interface. If no IPv4 addressing exists, then you must manually configure the router ID for IPv6 with the following command:
bgp router-id <ipv4 address>
Stuck-in-Active: Journal of an IT-Network Administrator 