Cannot disconnect or unregister inaccessible VMs from a "Not Responding" ESXi host following hardware failure

Products

• VMware vCenter Server
• VMware vSphere ESXi

Issue/Introduction

In VMware vCenter Server, an ESXi host remains in a Not Responding state following a hardware failure or network isolation. Administrators are unable to perform standard lifecycle operations on the host (even through ILOM) or its associated virtual machines. 

• Symptoms:

  • Virtual Machines appear as Inaccessible or Orphaned.

  • The "Disconnect" option for the host is greyed out or fails with a timeout.

  • The "Remove from Inventory" option for VMs is unavailable because vCenter still perceives the VMs as "Powered On."

Environment

 

• VMware vCenter Server 6.x / 7.x / 8.x
• ESXi 6.x / 7.x / 8.x

 

Cause

The vCenter Server service (vpxd) cannot complete a "Disconnect" handshake with the host-side management agent (vpxa) because the host is unreachable. Because the host is not "Disconnected," the inventory objects (VMs) remain locked in the vCenter database to prevent data corruption or split-brain scenarios.

Resolution

If the host is permanently failed and cannot be brought back online, follow these steps to clear the inventory:

• Force Remove the Host:

  1. In the vSphere Client, right-click the Not Responding host.

  2. Select Remove from Inventory.

  3. Click Yes to confirm the removal.

Note :  This action bypasses the disconnect requirement and purges the host and its registered VMs from the vCenter database. Removing the host from vCenter inventory does not send power commands to the underlying hypervisor; any surviving virtual machines running on the isolated host will continue to run unmanaged. However, VM configuration files (.vmx)  are allowed on shared storage to be re-registered to healthy hosts in the cluster via PowerCLI or the vSphere Client

• Restart vCenter Services (If UI is unresponsive): If the removal task hangs or the UI does not update, restart the vpxd service to clear stale sessions 

Note :Restarting vmware-vpxd will temporarily disconnect all administrators, drop all active vSphere Client sessions, and halt active cluster operations (such as ongoing storage migrations or scheduled tasks) for the duration of the initialization. It does not affect the data path or runtime state of virtual machines running on healthy hosts 

• 1. SSH to the VCSA as root.

  2. Run: service-control --restart vmware-vpxd

• Restore VM Connectivity: If the VM disks are on shared storage (vSAN, NFS, or iSCSI/FC):

1. 1. Navigate to the Storage view.

   2. Right-click the datastore > Files.

   3. Locate the VM folder and the .vmx file.

   4. Right-click the .vmx and select Register VM.