NSX Edge VMs are not restarted after the power outage of the Datacenter

Products

• VMware NSX
• VMware vSphere ESX 8.x
• VMware vCenter Server 8.0

Issue/Introduction

• In a real or simulated full power outage of a DC, after power is restored, all ESXi hosts and appliances (vCenter, HCX, NSX managers) start correctly, but NSX Edges do not.
• Logs show that HA cannot restart the Edge VMs due to network "none" being unavailable even though the network is configured to "start connected=FALSE". Following log lines seen in Hostd.log:

Wa(164) Hostd[2107429]: [Originator@6876 sub=Vigor.Vmsvc.vm:/vmfs/volumes/vsan:################-################/3525c368-####-####-####-############/NSX-Edge-0.vmx opID=host-41:306:#########-0-d56d sid=######## user=vpxuser] Unable to get network object for network none

Environment

• VMware vCenter Server 8.0.3
• VMware vSphere ESXi: 8.0.3
• VMware NSX: 4.2.2.1

Cause

• NSX manager connects unused uplinks to a special opaque 'none' port-group. On vCenter port-group 'none' is not accessible to user and cannot be assigned manually.
• vCenter ignores disconnected network on power on, and HA when vCenter is available ignores disconnected network during HA event as well.
• When HA event occurs while vCenter is down (which is the case when there is global power outage and everything is doing dark start) the HA does not ignore the disconnected network, and fails to power ON Edge VM as it unable to find network "none" in vCenter.

Resolution

• Removing the unused Edge interface from VM data path would allow HA to proceed with power ON of Edge VMs without looking for the opaque network "none" associated with unused uplinks by default.
• For steps to remove unused Edge VM interfaces, refer to the following KB: Edge NIC link status down alarm