VIDB upgrade from 9.0 to 9.1 fails with platform health check error platform-machines-core

Products

• VMware SDDC Manager / VCF Installer

Issue/Introduction

• Upgrade of VIDB from version 9.0 to 9.1 fails during the platform health check with the following error:
  
  Platform-machines-core: 1 of 4 resources are not ok
  
  
• This issue typically occurs after attempting to resolve datastore inconsistencies by running a datastore reclaim script Error: "Source and target systems use different datastores" - Upgrade of VCF Components from 9.0 to 9.1 Blocked by Datastore Mismatch Precheck Failure.
• The target datastore is part of a datastore cluster, and the datastore information was incorrectly entered in below format:

/<DatacenterName>/datastore/<DatastoreName>.
  
  The correct path format of a datastore should be:
  
/<DatacenterName>/datastore/<DatastoreClusterName>/<DatastoreName>


• Execution of kubectl get machines -A on the VIDB platform VM indicates a resource deployment state locked in the provisioning.
• Target virtual machines fail to appear inside the vCenter Server inventory.
• Below commands display an invalid identity structure containing an incorrect datastore directory path: /<DatacenterName>/datastore/<DatastoreName>

kubectl describe machine <machine_name>
kubectl describe vspheremachinetemplate <vspheremachinetemplate_name>

Environment

VMware Cloud Foundation 9.0.x
VMware Cloud Foundation 9.1.0

Cause

The vspheremachinetemplate asset definition contains a malformed datastore path string that prevents the automated initialization of subsequent compute objects.

Resolution

1. Confirm the presence of a verified, valid backup copy of the core architecture configuration state before executing modification tasks.
2. Open a secure shell (SSH) session to the VIDB platform management node using root credentials.
3. Correct the target data value mappings inside the package deployment structure:
   
   kubectl patch packagedeployment vmsp-platform -n vmsp-platform --type='merge' -p '{"spec":{"values":{"provider":{"vsphere":{"datastore":"NEW_DATASTORE_NAME"}}}}}'


4. Perform structural repairs on the vspheremachinetemplate
   • Isolate the precise runtime template identifier string:
     
     kubectl get vspheremachinetemplates -A


   • Extract data structures directly into a local storage file
     
     kubectl get vspheremachinetemplate <vcf-mgmt-####> -n vmsp-platform -o yaml > machine-temp.yaml


   • Modify the exported text layout file to align the datastore mapping values with the required path declaration syntax.
     
     /<DatacenterName>/datastore/<DatastoreClusterName>/<DatastoreName>
     
     
   • Delete the existing template
     
     kubectl delete vspheremachinetemplate <vcf-mgmt-####> -n vmsp-platform
     
     
   • Recreate vspheremachinetemplate with datastore path updated yaml file
     
     kubectl apply -f machine-temp.yaml
     
     
   • Monitor the VIDB platform control loop as the orchestration engine initializes machine redeployment
     
     kubectl get machines -A


   • Verify that the spec definitions inside the target cluster deployment records contain the correct datastore path layout by executing the following status trace commands:
     
     kubectl describe machine <MACHINE_NAME>
kubectl describe vspheremachinetemplate <VSPHEREMACHINETEMPLATE_NAME>