VCF 9.0.1 Installer fails at "Deploy vCenter" with error "Failed to query ESA certified disks" due to stale vSAN COMPUTE mode.

Products

VMware SDDC Manager VMware Cloud Foundation

Issue/Introduction

During the Bring-up process of VMware Cloud Foundation (VCF) 9.0.x, the deployment fails during the Deploy vCenter stage. The VCF Installer UI reports a failure to connect to the ESXi host, even though SSH connectivity and root credentials are confirmed to be working.

Error in VCF Installer UI:

Failed to connect to ESXi esxi01.example.com with user root Reference Token: #####

Error in /var/log/vcf/domainmanager/domainmanager.log in VCF Installer

ERROR [vcf_dm,697................................] [c.v.v.v.f.a.GenerateDeployVcenterInputAction,dm-exec-30]  An error occurred while generating the vsan deployment input
com.vmware.evo.sddc.common.client.vmware.vsphere.VsphereException: Failed to query ESA certified disks.

ERROR [vcf_dm,697................................] [c.v.e.s.o.model.error.ErrorFactory,dm-exec-30]  [######] VSPHERE_SDK_CONNECTING_HOST_FAILED Failed to connect to ESXi esxi01.example.com with user root
com.vmware.evo.sddc.orchestrator.exceptions.OrchTaskException: Failed to connect to ESXi esxi01.example.com with user root
        
Caused by: com.vmware.evo.sddc.common.client.vmware.vsphere.VsphereException: Failed to query ESA certified disks.

Caused by: com.vmware.evo.sddc.common.client.vmware.vsphere.VsphereOperationException: Task: (MOR:haTask--vim.host.VsanSystem.update-####) (Name:update) Entity: (MOR:ha-host) (Name:ha-host) is failed

In ESX /var/run/log/hostd.log it reports that vSan mode is set to Mode_Compute

In(14) localcli[4#####0]: VsanInfoCluster: vSan mode is set to : Mode_Compute

In ESX /var/run/log/vsansystem.log it reprots that vSAN personality is already set with a personality different that the requested personality

Er(163) vsansystem[2#####1]: [vSAN@#### sub=Libs opId=cf0#######] VsanInfoImpl: Error validating inputs: Not allow to enable vSAN if vSAN personality is already set with a personality 4 different than the requested personality - 1
Er(163) vsansystem[2#####1]: [vSAN@#### sub=VsanSystemProvider opId=cf0#######] Error validating vSAN inputs: Internal error occured, Please retry later. If the issue still persists, please reboot the host and retry.
Er(163) vsansystem[2#####1]: [vSAN@#### sub=VsanSystem opId=cf0#######] Error occurs when invoking Update with unexpected fault: N7Vmacore9ExceptionE(Internal error occured, Please retry later. If the issue still persists, please reboot the host and retry.)

Environment

VMware Cloud Foundation

Cause

The issue is caused by an inconsistent state in the ESXi host's vSAN stack.

The host has a stale "Personality" (Mode) set to COMPUTE (Personality 4), likely from a previous partial deployment or factory configuration.
VCF 9.0.x requires the host to be in Standard/HCI mode (Personality 1) for the Management Domain.
Because the personality is already locked to COMPUTE, the hostd/vSAN service rejects the VCF Installer's request to query ESA-certified disks, leading to a connection timeout/failure in the SDDC Manager logic.

Resolution

To resolve this, the stale vSAN cluster metadata must be cleared manually from the affected ESXi host(s) to allow the VCF Installer to reset the personality.

Steps to follow:

Verify the Inconsistent State

Log in to the affected ESXi host(s) via SSH and run:
```
esxcli vsan cluster get
```

Expected Result (In Error State): Mode: COMPUTE and a Sub-Cluster UUID will be present, even if Enabled: false.

Sample output

Cluster Information:
   Enabled: false
   Current Local Time:
   Local Node UUID:
   Local Node Type:
   Local Node State:
   Local Node Health State:
   Sub-Cluster Master UUID:
   Sub-Cluster Backup UUID:
   Sub-Cluster UUID: 7e98###########9d
   Sub-Cluster Membership Entry Revision: 0
   Sub-Cluster Member Count: 0
   Sub-Cluster Member UUIDs:
   Sub-Cluster Member HostNames:
   Sub-Cluster Membership UUID:
   Unicast Mode Enabled: true
   Maintenance Mode State:
   Config Generation:
   Mode: COMPUTE
   vSAN ESA Enabled: false
   vSAN Max Client Network Enabled: false

Clear Stale Metadata: Run the following commands to strip the stale personality:
1. Leave the ghost vsan cluster:
```
esxcli vsan cluster leave
```
2. Clear unicast/network metadata:
```
esxcli vsan network clear
```
Verify Reset: Run the get command again:
```
esxcli vsan cluster get
```
The Mode should no longer show COMPUTE, and the Sub-Cluster UUID should be empty.
Reboot and Retry
1. Reboot the ESXi host to ensure the VsanSystem service initializes with a clean state.
2. Go back to the VCF Installer UI and click Retry.

Note: If the issue persists after these steps, the host may have persistent partition metadata. A clean re-installation of ESXi 9.0.x is recommended for the affected nodes