Failed Deletion of NSX Compute Manager Due to Stale Edge Node References

Products

• VMware NSX

Issue/Introduction

When attempting to delete an old or unused Compute Manager from NSX-T. The deletion fails, preventing the removal of decommissioned vCenter instances from the NSX inventory. This typically occurs even after associated components like Edge VMs and NSX Managers have been migrated or re-deployed.

Environment

VMware NSX-T Data center
VMware NSX

Cause

The primary cause of the Compute Manager deletion failure is the persistent association of NSX Edge Node and NSX Manager entries with the old, decommissioned vCenter within the NSX-T inventory. Even if the Edge VMs and Managers have been re-deployed or moved to a new Compute Manager, NSX-T's internal mapping may still reference their original registration under the old vCenter's UUID.

Migration Incompleteness: While Edge VMs and NSX Managers might appear to be functioning under a new Compute Manager, their underlying inventory records in NSX-T still point to the old vCenter.

Decommissioned vCenter: The old vCenter being offline or decommissioned prevents NSX-T from performing a clean synchronization or de-registration of these stale entries.
API Confirmation: Internal API calls confirm that Edge Node and NSX Manager objects retain a link to the UUID of the old, decommissioned vCenter, preventing the associated Compute Manager from being successfully removed.

Resolution

Re-deploy Edge Nodes:

If not already done, ensure that all critical Edge Node services are migrated or re-deployed to function under the new Compute Manager.
This typically involves deploying new Edge Nodes and configuring them, then migrating services (like Tier-0/Tier-1 gateways) to them.
This step ensures that new, clean entries for the Edge Nodes are registered with the active Compute Manager.

Once the new Edge Nodes are operational and services migrated, any entries for the old Edge Nodes that were linked to the decommissioned vCenter must be explicitly removed. This is crucial as their stale presence prevents Compute Manager deletion.
Use an API client (e.g., Postman, curl) to execute the following NSX-T API call. This API action specifically targets and removes stale transport node entries (which include Edge Nodes) that are no longer accessible by the NSX Manager.

POST https://<nsx-mgr-ip-or-fqdn>/api/v1/transport-nodes?action=clean_stale_entries

Upon successful execution, the API will remove any transport nodes (including Edge Nodes) that are marked as stale.

After running the API call, verify that the stale Edge Node entries no longer appear in the NSX-T UI or via API queries.

Once the stale Edge Node entries are successfully cleaned up, retry the deletion of the problematic Compute Manager from the NSX-T UI or via API.
The Compute Manager should now be successfully deleted.

Additional Information

For further reference on cleaning up stale Edge Node entries, please refer to the following Broadcom Knowledge Base article:

https://knowledge.broadcom.com/external/article/345848/cleanup-stale-edge-nodes-present-in-the.html