NSX Upgrade Failure Due to Stale Transport Nodes Linked to Decommissioned Compute Manager

Products

• VMware NSX

Issue/Introduction

• NSX upgrade (e.g., from 4.2.2 to 4.2.3.x) fails during pre-check or host upgrade stage.
• Pre-upgrade check will fail with the following "Error while checking free space on tmp partition on host <Hostname / UUID>. Check the configuration state of host and ensure that services are operating normally."
• The NSX Manager UI displays stale Transport Nodes in an "Orphaned" or "Disconnected" state.
• Hosts are missing from the vCenter Server inventory but remain in NSX.
• Attempting to "Delete NSX" with the Force option or using standard API DELETE calls (Options 1–4 in KB# 319975) fails to remove the entries.

Environment

VMware NSX 4.x

VMware Cloud Foundation (VCF)

Cause

The ESXi host entries remain in the Corfu database because they maintain a dependency on a decommissioned or unreachable vCenter Server that is still registered as a Compute Manager in NSX. Even if the hosts are removed from vCenter, the stale Compute Manager registration blocks the final deletion of the transport node objects.

Resolution

To clear stale entries in existing versions: 

1. Identify Orphaned Compute Managers:
   • Navigate to System > Fabric > Compute Managers.
   • Check for any vCenter registrations that are in a "Down," "Disconnected," or "Unreachable" state and are known to be decommissioned.
     
     
2. Remove the Stale Registration:
   • Select the decommissioned vCenter and click Delete.
   • Wait several minutes for the deletion to sync across the management plane.
     
     
3. Verify Host Removal:
   • Navigate to System > Fabric > Nodes > Host Transport Nodes.
   • Check if the stale ESXi hosts have been automatically cleared from the inventory.
   • If they persist, try the Force Delete option from the UI again.
   • Database Resync (If required):
   • If the hosts still appear, SSH into the NSX Manager CLI as 'admin', and run:
start search resync all
   • The resync should be run on all Managers across the cluster. 

 

      4. Resume Upgrade:

• • Rerun the NSX upgrade pre-checks. The "node already exists" or "upgrade units" errors should be cleared.

 

For defects and enhancements, subscribe to this article (reference https://knowledge.broadcom.com/external/article/275360) to stay updated. 

Additional Information

To speak with a customer representative or a Support Engineer see Contact Support and scroll to the bottom of the page and click on your respective region.