EVPN routes are filtered if the encapsulation type in the route differs from the encapsulation type configured for the neighbor.
The following table summarizes the next-hop address set in the EVPN MP_REACH NLRI in BGP updates originated by a router irrespective of eBGP/iBGP peering type.
| MPLS | VxLAN |
|---|---|
|
BGP peering IP address |
Source VTEP IP address |
When there are intermediate BGP routers between the service end points, and these routers do not participate in tunnel termination and reorigination without routing in the overlay, BGP neighbors on the intermediate routers should be configured to keep the next-hop unchanged. This requirement applies to all types of tunnels, including MPLS VxLAN.
Support for IPv6 overlay next-hops is not available. The next-hop value in the EVPN routes is "IPv4 address," irrespective of IPv4 or IPv6 BGP peering.
The edge/leaf router performs next-hop resolution according to the encapsulation type. In the case of VxLAN, if VTEP discovery is enabled, VxLAN tunnels are automatically created and destroyed. For MPLS, tunnels are discovered by means of LDP/RSVP protocols. BGP resolves the EVPN MPLS next-hop with the MPLS tunnel, and routes are installed only when the next-hop is reachable. In case the MPLS tunnel becomes unreachable, the EVPN next-hop is marked as unresolved and routes are removed from the system.
The inclusive multicast route is originated for each VLAN/BD configured under EVPN, and this route triggers the extension of a corresponding VLAN/BD on the tunnel at the remote end. Import rules for inclusive multicast routes are discussed in BGP routing tables. Inclusive-multicast routes are required to carry the P-Multicast Service Interface (PMSI) tunnel attribute (RFC 6514, Sec. 5) in the path-attribute to signal the tunnel attributes. PMSI tunnel attribute fields are filled as shown in the following table.
| Attribute Field | Description |
|---|---|
|
Tunnel Type |
Tunnel type is always "Ingress Replication" (value 6) for all types of tunnels. No other type of tunnel is supported currently. |
|
MPLS label |
This field carries the L2 VNI for VxLAN and the EVI label for MPLS. |
|
Tunnel Identifier |
This field carries the source VTEP IP address for VxLAN and the BGP peering IP address for MPLS. |
In the case of LVTEP, both of the MCT cluster members advertise inclusive-multicast routes with different router IDs but the same next-hop. VLAN/BD membership of the tunnel is removed only when both advertising routers withdraw their inclusive-multicast routes. Therefore, consistent VLAN/BD configuration is required in an MCT cluster.
In the case of MPLS, there is always one tunnel between the MCT peers. In addition, inclusive-multicast route affects only the tunnel from which the route is originated. In order to perform load balancing of the unicast traffic, aliasing is used as discussed in the next section.