The NSX-T Edge node receives EVPN Type-5 routes that advertise the same VTEP IP but with different RMAC values from two different TOR switches.
The frr.log shows events indicating RMAC changes for the same remote VTEP IP.
2025/09/14 04:16:06.692985 ZEBRA: L3VNI <VNI> RMAC change (old rmac --> new rmac) for nexthop 1.#.#.#, prefix 2.#.#.#/322025/09/14 04:16:06.693035 ZEBRA: RMAC <new rmac> added L3VNI <VNI-ID> Remote VTEP 1.#.#.#
VMware NSX-T Data center 3.x
VMware NSX
This issue is due to a misconfiguration on the Border Leaf (BL) switches.
Initially, the NSX-T Edge node received identical EVPN Type-5 routes from both BL1 and BL2. BGP selected the route from BL1 as the best path.
Later, the Edge node again received Type-5 routes from both BL1 and BL2, but this time the route from BL2 was preferred. As a result, the Edge updated the forwarding entry and removed BL1’s RMAC, causing traffic destined for the shared VTEP IP (advertised by both BL1 and BL2) to be dropped.
This behavior occurs because BL1 and BL2 advertise Type-5 routes using the same VTEP IP but different RMAC values, which leads to RMAC flapping and forwarding inconsistencies on the Edge node.
Use one of the following configuration options on the Border Leaf switches:
Option 1:
Configure BL1 and BL2 to advertise EVPN Type-5 routes using the same VTEP IP (VIP) and the same RMAC (VMAC).
Option 2:
Configure BL1 and BL2 to advertise EVPN Type-5 routes using different VTEP IPs (PIP) and different RMAC values (PMAC or SYSMAC).