You are planning to use distributed routing for overlay subnets in an EVPN VXLAN fabric. The customer needs to provide DHCP addresses to some VMs in the tenant-a VRF. The DHCP server is in the…

Question

Options

• AChange the design to use centralized routing because HPE Aruba CX switches cannot support
• BFor every SVI, assign a unique actual IP address, different from the active gateway IP address
• COn each VTEP switch, create a loopback interface in the shared VRF with the same IP on both
• DOn each VTEP switch, create a loopback interface in the tenant-a VRF with a different IP on both

• A
  27% (16)
• B
  8% (5)
• C
  50% (30)
• D
  15% (9)

Explanation

Option C is correct because inter-VRF DHCP relay in a distributed EVPN VXLAN fabric requires the relay agent to have a reachable source IP in the VRF where the DHCP server resides (the shared VRF). Creating a loopback in the shared VRF with the same IP on every VTEP creates an anycast address - the DHCP server's reply travels back to that IP, and EVPN/VXLAN routes it to whichever VTEP the client is currently attached to, making the relay work transparently regardless of client placement.

Why the distractors are wrong:

• A is false - HPE Aruba CX switches fully support inter-VRF DHCP relay with distributed routing; changing to centralized routing is unnecessary and introduces a bottleneck.
• B describes assigning unique per-VTEP IPs to SVIs alongside the anycast gateway IP, which is a valid anycast gateway concept but does nothing to enable DHCP relay across VRFs.
• D puts the loopback in the wrong VRF (tenant-a instead of shared) and uses different IPs per VTEP - breaking the anycast behavior the DHCP server needs to send replies back to the relay agent.

Memory tip: "Same loopback, shared VRF" - the loopback lives where the server lives (shared VRF), and the same IP on all VTEPs mirrors how anycast gateways work, letting the server's reply find its way home to any VTEP.

No community discussion yet for this question.