Overlay VM network connectivity loss with NSX Multi-TEP

Products

• VMware NSX

Issue/Introduction

VMs on an NSX overlay segment lose network connectivity after migrating (vMotion) to hosts within a cluster. Connectivity is restored only when one of the two uplinks is in unused.

• NSX reports "faulty TEP" on affected hosts.

• vmkernel interface is up vmkping between TEP interfaces succeeds, but overlay traffic fails.

vmkping -I vmkx (IP)-S vxlan 

• nsxdp-cli bfd sessions list shows tunnels are up, but Layer 2 forwarding fails
  .
• The issue occurs when the vSphere Distributed Switch (VDS) uses two active-active uplinks.
  
  

Environment

• VMware NSX 4.x

• VMware vSphere 7.x / 8.x

Cause

The upstream physical switch interfaces are configured for a Port Channel (LACP), which conflicts with the NSX Multi-TEP active-active configuration. NSX Multi-TEP requires that the physical switch ports remain as independent trunk ports to handle the multiple TEP interfaces correctly.

Resolution

This issue may occur due to an invalid configuration—like having LACP enabled on the physical switch ports while using an Active/Active uplink policy in vDS teaming

1. Verify the physical switch configuration for the ports connected to the ESXi host uplinks.

2. Remove the LAG (Port Channel/EtherChannel) configuration on the physical switch.

3. Configure the physical switch ports as independent trunk ports.

4. Modify the NSX Uplink Profile to use standard uplinks (e.g., uplink1, uplink2) in an active-active teaming policy.

5. Ensure the physical NICs (e.g., vmnic2, vmnic3) are assigned to the correct VDS uplinks.

6. Migrate a test VM to verify stable connectivity across all hosts in the cluster.

Additional Information

For detailed configuration requirements, see Configure an ESXi Host Transport Node with Link Aggregation Group and also, Configuring a LAG on a vSphere Distributed Switch Port Group when using LACP