NSX Federation: On 3.0 and 3.1, cross-site connectivity for a stretched T1 is lost upon putting an Edge into Maintenance Mode if its Edge cluster originally had only a single Edge node

Products

• VMware NSX

Issue/Introduction

Symptoms:
- NSX Federation is in use on versions 3.0.x or 3.1.x
- Cross-site connectivity is lost on stretched T1 DR or SR when Active Edge is put into Maintenance Mode or powered off 
- The T1's Edge cluster originally had only a single edge node. At least one more Edge was added to the cluster later
- 'get logical-switch <Switch-UUID> l2forwarders' output shows only one Edge node

Cause

Only Active / Standby is supported for T1 L2 E/W Intersite Forwarding. Standby allocation of T1 SR l2 forwarders was not done automatically before.

Resolution

This issue is resolved in NSX-T 3.2 and later releases.

Workaround:
Edge clusters should have at least two Edge nodes to ensure high availability.

Additional Information

Impact/Risks:
Even though additional Edge nodes were later to the Edge cluster, there is no HA failover for cross-site communication if one Edge is put into Maintenance Mode.