This topology uses two types of BGP address families: IPv4 and EVPN. The leafs peer with the spines for both IPv4 and EVPN, as well as the DCI switches for IPv4 only. The leafs in Site 1 that are connected to the DCI switches peer with the leafs in Site 2 that are connected to the DCI switches for EVPN only. MC-LAG leaf pairs connected to DCI switches peer with each other for IPv4 and EVPN.
The table below summarizes the BGP peering configuration:
Switch | Peers with | Address family |
Leaf 1A and 1B | Spine 1 and Spine 2 | IPv4 and EVPN |
Leaf 2A and 2B | Spine 1 and Spine 2 | IPv4 and EVPN |
Leaf 2A and 2B | DCI Site 1 | IPv4 |
Leaf 2A and 2B | Leaf 3A and 3B | EVPN |
Leaf 2A and 2B | Leaf 2A peers with 2B; Leaf 2B peers with 2A | IPv4 and EVPN |
Leaf 3A and 3B | DCI Site 2 | IPv4 |
Leaf 3A and 3B | Leaf 2A and 2B | EVPN |
Leaf 3A and 3B | Leaf 3A peers with 3B; Leaf 3B peers with 3A | IPv4 and EVPN |
The BGP IPv4 address family is used to build traditional route tables for IPv4 addresses in the topology. These tables include paths to router IDs, VTEP addresses, and VxRail network addresses. You can determine the IPv4 peering status in SONiC with the show bgp ipv4 unicast summary
command. To view the route tables, run show ip route
for the default VRF and show ip route vrf vrfname
for the management and tenant VRFs.
BGP EVPN peers are used for VXLAN routing, which is sometimes referred to as routing in and out of tunnels (RIOT). The information exchanged includes virtual networks, MAC addresses, and their associated VTEP addresses. You can determine EVPN peering status in SONiC with the show bgp l2vpn evpn summary
command. To view information learned through BGP EVPN, Dell Enterprise SONiC includes several commands prefixed with show vxlan
and show evpn
. When EVPN is configured, the commands show ip arp
and show mac address-table
also include the router ID of the applicable VTEP for each MAC address learned.
In the switch configuration examples attached, the leaf switches peer with the spines and the DCI switches using unnumbered BGP. This peering method is optional but simplifies configuration. With unnumbered BGP, IPv6 is enabled on the ports connected to the peer using the command ipv6 enable
. BGP uses the automatically generated IPv6 link-local address to establish a BGP session with the neighbor. The command capability extended-nexthop
is used to enable unnumbered BGP. While the leaf-to-spine and leaf-to-DCI links are established using IPv6, the routes exchanged in this example are all IPv4.