If a vSphere distributed switch is removed from vCenter Server in an unsupported way, for example, hosts being forcibly removed from the distributed switch, or the vCenter Server database being reconstructed due to corruption, you might encounter an invalid device backing
error for the network interface of a virtual machine (vNIC).
Virtual machines or vNIC devices are pointing to a distributed switch or distributed port group that no longer exists. You might also not be able to remove the corresponding proxy switches on the hosts because the ports from the incorrectly removed distributed switch are still in use.
To resolve this issue, you must migrate all affected hosts and virtual machine NIC devices to a separate network. You can apply a valid configuration after the migration.
To migrate the affected virtual machines and their vNIC devices to a separate network, use the Migrate virtual machine networking wizard in the vSphere Client and vSphere Web Client, as described in the vSphere Networking Guide. Select all virtual machines listed with no networking – these are the virtual machines with the invalid device backing error. After the vNIC devices are migrated away from the invalid distributed switch, you can remove the proxy switch, or allow vCenter Server to do it automatically. vCenter Server automatically removes a proxy switch from a host when it detects a proxy switch without a corresponding distributed switch.
Select all virtual machines with no networking connectivity.
In the vSphere Client, after you move the host to the new switch with a valid uplink to carry traffic, the error disappears as soon as you switch virtual machine traffic to a port group that has uplink connectivity.
In vSphere 5.1 and later, to migrate the physical NICs, VMkernel adapters and virtual machines from an affected host to a separate network, for example on a new distributed switch, perform these steps in the vSphere Web Client:
Select all virtual machines with no networking connectivity.
In the vSphere Web Client, after you move the host to the new switch with a valid uplink to carry traffic, the error disappears as soon as you switch virtual machine traffic to a port group that has uplink connectivity.