Multicast traffic propagation fails while using stretched cluster

Products

• VMware NSX

Issue/Introduction

Multicast propagation fails in one or both directions in an environment configured as follows:
- Cisco UCS fabric and Nexus 9k switches upstream
- NSX 4.2.x configured with overlay networking
- VM connected to a overlay-backed segment uplinked to T1 which is uplinked to T0
- T0 configured to peer with 9k upstream using BGP
- BGP peers established and unicast communication North-South and East-West working correctly
- Multicast is enabled on the T1 router
- PIM routing is configured correctly on the T0
- Rendezvous point is correctly configured in the PIM profile
- The VM communicates via multicast to a sender/receiver in the physical underlay
- The VM lives in a vSphere cluster that is stretched across physical locations
- Cisco UCS Manager server profile for ESXi hosts is configured to not block PIM/Multicast join packets

Environment

VMware NSX 4.2.x

Cause

This happens because the vSphere stretched cluster does not have L2 multicast forwarding configured within it. For PIM forwarding to work correctly, the vSphere Cluster needs to be able to propagate multicast traffic between the MTEP and it's local TEPs across both sites via L2. 

Resolution

Enable L2 multicast forwarding in the physical underlay across the stretched cluster.

Additional Information

Configuring multicast in NSX:
Configuring Multicast on an NSX Tier-0 or Tier-1