Communication failure after HCX Network Extension appliance vMotion

Products

• VMware HCX

Issue/Introduction

• Communication over HCX network extension fails for a while after NE appliances vMotion.
• Eventually it recovers with no manual operation.
• The NE appliance is connected to vSphere portgroups, not NSX segments.

Environment

VMware HCX deployed on non-NSX environment.

Cause

After vMotion, packets over network extension needs to be forwarded to the new ESXi and the MAC address table entries on the physical switches need to be updated accordingly.
However, current HCX NE appliances do not recognize vMotion and do not send out any packets to update MAC address table.
Thus the MAC address table entries on the physical switches are stale until they expire or packets from the VMs over network extension arrive from the remote site.

Resolution

• Currently, there is no resolution.
  It is planned to be fixed in future releases so that RARP is sent out on vMotion to update MAC address table.

Additional Information

• To prevent the issue, refrain from manual vMotion and override DRS automation level to suppress vMotion by DRS.
  https://techdocs.broadcom.com/us/en/vmware-cis/vsphere/vsphere/8-0/vsphere-resource-management-8-0/creating-a-drs-cluster.html#GUID-C21C0609-923B-46FB-920C-887F00DBCAB9-en
  • Note: You need to configure it again after redeploy and upgrade.
• In case of ESXi maintenance, configure HA and shutdown the NE appliances rather than vMotion.
• It recovers communication over L2 extension immediately to send some packets from the remote site VMs to the local site.