IPv6 and Address-Family Support in BGP
IPv6 and Address-Family Support in BGP
This domain tests MP-BGP (RFC 4760) and the SR OS family configuration that enables IPv6 NLRI exchange.
Multiprotocol extensions
Classic BGP-4 carried only IPv4 unicast. MP-BGP adds AFI/SAFI identifiers in the OPEN capability and replaces the IPv4-specific NLRI fields with MP_REACH_NLRI and MP_UNREACH_NLRI attributes. IPv6 unicast is AFI 2, SAFI 1. Because the family is negotiated in the OPEN, both sides must enable it. A session can establish for IPv4 while IPv6 routes never flow.
In SR OS, family ipv4 ipv6 under a group or neighbor enables both AFI/SAFIs, and a single TCP session can carry both families between dual-stack peers. If the transport is IPv4 but you want IPv6 routes, you still enable family ipv6 on that IPv4 neighbor.
Next hop: for IPv6 routes the next hop is an IPv6 address, usually the peer's global address or a link-local address. A link-local next hop requires the receiving router to know the outgoing interface, so link-local peering is normally confined to directly connected eBGP. Extended next hop (RFC 5549) lets an IPv4 route carry an IPv6 next hop, which is useful in IPv6-only underlay designs.
Worked example
Two dual-stack routers peer over IPv4 192.0.2.1 and 192.0.2.2 and must exchange both tables. Configure family ipv4 ipv6 on both neighbors and an export policy that admits the IPv6 prefixes. Verify with show router bgp neighbor 192.0.2.2, which lists the negotiated families, and show router bgp routes ipv6.
Traps
Assuming an Established IPv4 session means IPv6 works; enabling family ipv6 on only one side; expecting link-local next hops to work across multiple hops; and mixing up AFI/SAFI - AFI 2 is IPv6, SAFI 1 is unicast.