vGPU Driver download fails in DLVM 9.0.0.0

Products

VCF Private AI Services

Issue/Introduction

During online deployment, DLVM 9.0.0.0 fails to retrieve the vGPU driver catalog manifest file.

The following errors are logged in /var/log/vgpu-install.log:

YYYY-MM-DD 02:06:17,758 - dlvminit.api.url_content - get_content[ERROR]: Unexpected error while fetching content from https://packages.vmware.com/dl-vm/vgpu-ngc-manifest.txt
requests.exceptions.HTTPError: 404 Client Error: Not Found for url: https://packages.vmware.com/dl-vm/vgpu-ngc-manifest.txt
YYYY-MM-DD 02:06:17,761 - dlvminit.vgpu_url - parse_vgpu_driver_info[ERROR]: Failed to fetch content from URL: https://packages.vmware.com/dl-vm/vgpu-ngc-manifest.txt
...
YYYY-MM-DD 02:06:18,224 - dlvminit.logging_config - error[ERROR]: ERROR Driver installation failed

Environment

VMware Private AI Foundation 9.0.0.0 (DLVM 9.0.0.0)

Cause

Starting from October 2025, online deployments of DLVM 9.0.0.0 may fail to download vGPU drivers.

The vGPU driver catalog file used by DLVM 9.0.0.0 is no longer accessible as the depot, packages.vmware.com has been retired.

Resolution

To resolve the issue, use DLVM 9.0.1.0 or later releases.

In DLVM 9.0.1.0, the vGPU driver catalog file and vGPU drivers are downloaded directly from NVIDIA NGC using the API key provided by the user.
This removes the dependency on the packages.vmware.com depot and ensures continued access to the latest vGPU guest drivers.

For environments that run DLVM 9.0.0.0, to work around the issue, manually host the vGPU driver catalog manifest file.

Please use the following steps to manually host the vGPU driver catalog manifest file,

Prepare the vGPU driver catalog file: Host a mapping file (for example, vgpu-ngc-manifest.txt) within the local environment.
This file provides the mapping between each vGPU host driver version and the corresponding guest driver download URL.
Set up a local web server: Upload the vGPU driver catalog file to a local server so that it can be accessed during DLVM deployment.
Configure the DLVM OVF parameter: Enter the vGPU driver catalog file URL in the DLVM 'vgpu-url' OVF parameter:
- When deploying DLVM via vCenter Server, specify the URL in the 'Custom URL for vGPU Driver' field.
- When deploying DLVM via VCFA, specify the URL in the 'Local HTTP Server' field under the 'NVIDIA vGPU guest driver location' section.
- When deploying DLVM via VMservice, specify the URL in the 'vgpu-url' field.

For detailed instructions, please refer to VMware Private AI Foundation with NVIDIA 9.0 under the section 'Provide a location to download the vGPU guest drivers'.

Additional Information

VMware Private AI Foundation with NVIDIA 9.0