VIO instance stuck in deleting state when underlying VM in vCenter is inaccessible

Products

• VMware Integrated OpenStack

Issue/Introduction

• A virtual machine managed by VIO becomes inaccessible in vCenter (e.g., following a VM hardware version downgrade).
• Subsequent attempts to delete the VM or its associated volumes via the VIO CLI or UI fail.
• The nova-compute logs indicate that the resource tracker continuously attempts to lock and validate the instance, preventing its purge from the Placement service and Nova database.
  
  controller-ABC/nova/nova-compute.log

Feb 02 03:52:49 controller-ABC nova-compute[880]: 2026-02-02 03:52:49.671 1 DEBUG nova.compute.resource_tracker [req-xxxxxx-xxxx-xxxx-xxxx-xxxxxxxxx- - - - -] Instance xxxxx-xxxx-xxx-xxx-xxxx actively managed on this compute host and has allocations in placement: {'resources': {'DISK_GB': 100, 'MEMORY_MB': 32768, 'VCPU': 4}}. _remove_deleted_instances_allocations /usr/lib/python3.7/site-packages/nova/compute/resource_tracker.py:1591

         /var/log/vmware/vpxd.log 

2026-02-06T03:42:14.733Z error vpxd[08081] [Originator@6876 sub=StateLock opID=n-cpu-xxxxxx-xxx-xxx-xxxx-xxxxxx-xx] VM svcabcdefegh (xxxxxx-xxxx-xxx-xxx-xxxxx): Current state: inaccessible, IsConnected = false, expected = true, cryptoLocked = false

2026-02-06T04:47:39.201Z error vpxd[08987] [Originator@6876 sub=StateLock opID=n-cpu-xxxxxx-xxx-xxx-xxxx-xxxxxx-xx] VM svcabcdefegh (xxxxx-xxxx-xxx-xxx-xxxx): Current state: inaccessible, IsConnected = false, expected = true, cryptoLocked = false

Environment

• VMware Integrated OpenStack (VIO) 7.x
• VMware vCenter Server 7.0.x

Cause

• A state mismatch exists between the OpenStack Nova database and the vCenter inventory.
• When a VM enters an inaccessible state in vCenter, standard API-driven VM lifecycle operations are locked.
• The OpenStack Nova service issues a destroy API call, but vCenter rejects it.
• This leaves the task stuck in a deleting loop within VIO, as the compute resource tracker maintains the active allocation.

Resolution

To resolve the issue, the orphaned VM record must be cleared at the vSphere layer before resetting the OpenStack task state.

1. Ensure valid backups and snapshots of the vCenter Server and VIO management nodes are available.

2. In the vSphere Client, locate the inaccessible VM in the inventory .

3. Right-click the VM and select Remove from Inventory. Do not select "Delete from Disk".

4. Log in to the VIO controller node via SSH.

5. Reset the OpenStack instance state to active to break the hung deletion task: openstack server set --state active <INSTANCE_UUID>

6. Re-attempt the deletion of the instance via the CLI: openstack server delete <INSTANCE_UUID>