NVIDIA Virtual GPU (vGPU) Software

https://docs.nvidia.com/vgpu/
Admin
NVIDIA Virtual GPU (vGPU) Software provides virtualized GPU capabilities for NVIDIA RTX Virtual...
Tokens:1,019,361
Snippets:5,020
Trust Score:10
Update:11 months ago
Show doc for...
Tokens:
Raw
### Example PowerShell Output for NVIDIA Control Panel Installation

Source: https://docs.nvidia.com/vgpu/18.0/grid-vgpu-release-notes-microsoft-windows-server/index

This is an example output from the `Get-AppxPackage -AllUsers -Name NVIDIACorp.NVIDIAControlPanel` PowerShell command, showing the details of the installed NVIDIA Control Panel package, including its publisher, version, and installation status for different users.

```powershell
PS C:\> Get-AppxPackage -AllUsers -Name NVIDIACorp.NVIDIAControlPanel

Name                   : NVIDIACorp.NVIDIAControlPanel
Publisher              : CN=D6816951-877F-493B-B4EE-41AB9419C326
Architecture           : X64
ResourceId             : 
Version                : 8.1.964.0
PackageFullName        : NVIDIACorp.NVIDIAControlPanel_8.1.964.0_x64__56jybvy8sckqj
InstallLocation        : C:\Program Files\WindowsApps\NVIDIACorp.NVIDIAControlPanel_8.1.964.0_x64__56jybvy8sckqj
IsFramework            : False
PackageFamilyName      : NVIDIACorp.NVIDIAControlPanel_56jybvy8sckqj
PublisherId            : 56jybvy8sckqj
PackageUserInformation : {S-1-12-1-530092550-1307989247-1105462437-500 [Administrator]: Installed,
                         S-1-12-1-530092550-1307989247-1105462437-1002 [pliny]: Installed,
                         S-1-12-1-530092550-1307989247-1105462437-1003 [trajan]: Installed}
IsResourcePackage      : False
IsBundle               : False
IsDevelopmentMode      : False
NonRemovable           : False
IsPartiallyStaged      : False
SignatureKind          : Store
Status                 : Ok
```

--------------------------------

### Example Output of Get-AppxPackage -AllUsers

Source: https://docs.nvidia.com/vgpu/18.0/grid-vgpu-release-notes-microsoft-azure-stack-hci/index

This is an example output demonstrating the detailed information returned by the `Get-AppxPackage -AllUsers -Name NVIDIACorp.NVIDIAControlPanel` command. It shows the package name, publisher, version, installation location, and user-specific installation status.

```powershell
PS C:\> Get-AppxPackage -AllUsers -Name NVIDIACorp.NVIDIAControlPanel

Name                   : NVIDIACorp.NVIDIAControlPanel
Publisher              : CN=D6816951-877F-493B-B4EE-41AB9419C326
Architecture           : X64
ResourceId             :
Version                : 8.1.964.0
PackageFullName        : NVIDIACorp.NVIDIAControlPanel_8.1.964.0_x64__56jybvy8sckqj
InstallLocation        : C:\Program Files\WindowsApps\NVIDIACorp.NVIDIAControlPanel_8.1.964.0_x64__56jybvy8sckqj
IsFramework            : False
PackageFamilyName      : NVIDIACorp.NVIDIAControlPanel_56jybvy8sckqj
PublisherId            : 56jybvy8sckqj
PackageUserInformation : {S-1-12-1-530092550-1307989247-1105462437-500 [Administrator]: Installed,
                         S-1-12-1-530092550-1307989247-1105462437-1002 [pliny]: Installed,
                         S-1-12-1-530092550-1307989247-1105462437-1003 [trajan]: Installed}
IsResourcePackage      : False
IsBundle               : False
IsDevelopmentMode      : False
NonRemovable           : False
IsPartiallyStaged      : False
SignatureKind          : Store
Status                 : Ok
```

--------------------------------

### Install NVIDIA vGPU Software

Source: https://docs.nvidia.com/vgpu/toolkits/fine-tunning-llms/latest/deployment

This snippet demonstrates how to make the NVIDIA vGPU driver package executable and then install it using dpkg. Ensure the filename matches your downloaded version.

```bash
$ sudo chmod +x nvidia-linux-grid-xxx_xxx.xx.xx_amd64.deb
$ sudo dpkg -i nvidia-linux-grid-xxx_xxx.xx.xx_amd64.deb
```

--------------------------------

### Install NVIDIA vGPU Software

Source: https://docs.nvidia.com/vgpu/toolkits/rag-application/latest/deployment

This code block shows how to make the NVIDIA vGPU software installer executable and then install it using dpkg. Replace 'xxx_xxx.xx.xx' with the actual version number of the driver.

```bash
sudo chmod +x nvidia-linux-grid-xxx_xxx.xx.xx_amd64.deb
sudo dpkg -i nvidia-linux-grid-xxx_xxx.xx.xx_amd64.deb
```

--------------------------------

### XenServer NVIDIA GPU Manager Package Installation Example

Source: https://docs.nvidia.com/vgpu/4.9/grid-vgpu-user-guide/index

This example demonstrates verifying the installation of a specific NVIDIA vGPU Manager RPM package on Citrix XenServer. It shows the expected output when the package is installed and also illustrates potential conflict errors if an older version is already present and an upgrade is not explicitly selected.

```bash
[root@xenserver ~]# rpm –q NVIDIA-vGPU-xenserver-7.0-367.132
[root@xenserver ~]#
If an existing NVIDIA GRID package is already installed and you don’t select the upgrade (-U) option when installing a newer GRID package, the rpm command will return many conflict errors.
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-367.132.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.130.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-367.132.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.130.x86_64
        ...

```

--------------------------------

### Execute NIM Startup Script

Source: https://docs.nvidia.com/vgpu/toolkits/agentic-rag/latest/deployment

This command executes the previously created `start_ds.sh` script to download and start the NIM service. The process may take a significant amount of time depending on system resources.

```bash
./start_ds.sh
```

--------------------------------

### Install and Configure VMware Horizon Client

Source: https://docs.nvidia.com/vgpu/deployment/vmware/latest/index

Instructions for installing the VMware Horizon Client on end-user devices and configuring its connection to the Horizon Connection Server. This allows users to access their virtual desktops and applications.

```text
1. Download the VMware Horizon Client installer for your operating system (Windows, macOS, Linux, iOS, Android) from the VMware website.
2. Run the installer and follow the on-screen prompts to complete the installation.
3. Launch the VMware Horizon Client application.
4. Click 'New Server' or the '+' icon.
5. Enter the FQDN or IP address of your Horizon Connection Server.
6. Click 'Connect'.
7. You may be prompted to accept a certificate; click 'Connect' or 'Accept'.
8. Log in with your Active Directory credentials or other configured authentication method.
```

--------------------------------

### Example of XenServer Package Installation Conflict

Source: https://docs.nvidia.com/vgpu/4.5/grid-vgpu-user-guide/index

This example demonstrates the output when attempting to install a newer NVIDIA vGPU Manager package on Citrix XenServer without using the upgrade option, resulting in conflict errors due to file overlaps with an existing version. This helps in understanding and resolving installation problems.

```shell
[root@xenserver ~]# rpm –q NVIDIA-vGPU-xenserver-7.0-367.123
[root@xenserver ~]#
If an existing NVIDIA GRID package is already installed and you don’t select the upgrade (-U) option when installing a newer GRID package, the rpm command will return many conflict errors.
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-367.123.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.122.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-367.123.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.122.x86_64
        ...

```

--------------------------------

### Start WSL for NVIDIA Workbench on Windows

Source: https://docs.nvidia.com/vgpu/toolkits/agentic-rag/latest/deployment

This command is used on Windows guests to start the Windows Subsystem for Linux (WSL) with the NVIDIA Workbench distribution. It's a prerequisite step before running Docker commands related to NVIDIA AI Workbench on Windows.

```shell
wsl -u root -d NVIDIA-Workbench
```

--------------------------------

### Example NVIDIA NIM Log Output During Compose

Source: https://docs.nvidia.com/vgpu/toolkits/pdf-to-podcasts/latest/deployment

This is an example log output from a local NIM container during Docker Compose execution. It indicates the average prompt and generation throughput, request status, and KV cache usage.

```log
local-nim-1          | INFO 2025-03-14 20:48:40.382 metrics.py:351] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 0 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.0%, CPU KV cache usage: 0.0%.
```

--------------------------------

### Verify NVIDIA Driver and vGPU License with nvidia-smi

Source: https://docs.nvidia.com/vgpu/toolkits/agentic-rag/latest/deployment

This snippet demonstrates how to verify the NVIDIA guest driver operation and vGPU license status using the `nvidia-smi` command in a Windows command prompt or Linux console. This is crucial for confirming the successful installation and configuration of NVIDIA vGPU software.

```shell
nvidia-smi
```

```shell
nvidia-smi -q
```

--------------------------------

### Install NVIDIA vGPU RPM Package on XenServer

Source: https://docs.nvidia.com/vgpu/3.4/grid-vgpu-user-guide/index

Installs the NVIDIA vGPU RPM package on XenServer using the 'rpm -iv' command. This is the initial installation step. Requires the RPM file to be present on dom0.

```bash
[root@xenserver ~]# rpm -iv  NVIDIA-vGPU-xenserver-6.2-361.40.i386.rpm
Preparing packages for installation...
NVIDIA-vGPU-xenserver-6.2-361.40
[root@xenserver ~]#
```

--------------------------------

### Install NVIDIA vGPU RPM on XenServer

Source: https://docs.nvidia.com/vgpu/4.3/grid-vgpu-user-guide/index

Installs the NVIDIA vGPU RPM package on XenServer. This command is used for initial installation of the vGPU Manager. The package must be copied to dom0 prior to execution.

```bash
[root@xenserver ~]# rpm -iv NVIDIA-vGPU-xenserver-7.0-367.106.x86_64.rpm
Preparing packages for installation...
NVIDIA-vGPU-xenserver-7.0-367.106
[root@xenserver ~]#
```

--------------------------------

### Create and Configure NIM Startup Script

Source: https://docs.nvidia.com/vgpu/toolkits/agentic-rag/latest/deployment

This snippet covers the creation of a bash script for starting the NIM service. It includes commands to create the script file, set permissions, and define environment variables for cache and API keys. The script also handles directory creation, Docker login, and the execution of the NIM container. Dependencies include `bash`, `sudo`, `chmod`, `mkdir`, `docker`, and `curl`.

```bash
wsl -u root -d NVIDIA-Workbench
```

```bash
touch start_ds.sh
chmod ugo+rwx start_ds.sh
```

```bash
_#!/bin/bash_

_# Set the local NIM cache directory_
export LOCAL_NIM_CACHE=~/.cache/nim

_# Check if the NVIDIA cache directory exists_
if [ ! -d "$LOCAL_NIM_CACHE" ]; then
     _# Create the NVIDIA cache directory_
     sudo mkdir -p $LOCAL_NIM_CACHE
     echo "Created NVIDIA cache directory at$LOCAL_NIM_CACHE"
else
     echo "NVIDIA cache directory already exists at$LOCAL_NIM_CACHE"
fi

_# Set permissions on LOCAL_NIM_CACHE_
chmod -R a+rw "$LOCAL_NIM_CACHE"

_# Set the NGC API KEY_
export NGC_API_KEY='<your NVIDIA NGC API Key goes here>'

_# Log into docker/NGC_
docker login -u '$oauthtoken' --password-stdin nvcr.io <<< $NGC_API_KEY

_# Run the Docker container with the specified parameters_
docker run -it --rm \
     --gpus all \
     --shm-size=16GB \
     -e NGC_API_KEY \
     -v "$LOCAL_NIM_CACHE:/opt/nim/.cache" \
     -u $(id -u) \
     -p 8000:8000 \
     nvcr.io/nim/deepseek-ai/deepseek-r1-distill-llama-8b:1.5.2

_# Test if the nim is running_
docker ps
```

--------------------------------

### Example: Create A40-2Q vGPU Instance

Source: https://docs.nvidia.com/vgpu/11.0/grid-vgpu-user-guide

This example demonstrates creating an instance of the A40-2Q vGPU type by writing its specific UUID to the 'create' file within the 'nvidia-558' registration directory. This action registers the vGPU with the system.

```bash
# echo "aa618089-8b16-4d01-a136-25a0f3c73123" > nvidia-558/create
```

--------------------------------

### Example of NVIDIA GPU Manager Package Installation Conflict on XenServer

Source: https://docs.nvidia.com/vgpu/4.8/grid-vgpu-user-guide/index

This example demonstrates potential conflict errors when installing a newer NVIDIA GRID package on Citrix XenServer without using the upgrade option. Such conflicts indicate issues with package management and may require the removal of the older package.

```bash
[root@xenserver ~]# rpm –q NVIDIA-vGPU-xenserver-7.0-367.130
[root@xenserver ~]#
If an existing NVIDIA GRID package is already installed and you don’t select the upgrade (-U) option when installing a newer GRID package, the rpm command will return many conflict errors.
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-367.130.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.128.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-367.130.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.128.x86_64
        ...

```

--------------------------------

### Example gridd.conf for Quadro vDWS

Source: https://docs.nvidia.com/vgpu/10.0/grid-licensing-user-guide/index

An example of a gridd.conf file configured for Quadro vDWS, specifying license server addresses, ports, and the feature type. This serves as a template for users to adapt.

```properties
# /etc/nvidia/gridd.conf.template - Configuration file for NVIDIA Grid Daemon

# This is a template for the configuration file for NVIDIA Grid Daemon.
# For details on the file format, please refer to the nvidia-gridd(1)
# man page.

# Description: Set License Server Address
#ServerAddress=gridlicense1.example.com
#BackupServerAddress=gridlicense2.example.com
#ServerPort=7070
#BackupServerPort=7070
#FeatureType=2
```

--------------------------------

### Install NVIDIA vGPU Driver on Linux VM (Ubuntu)

Source: https://docs.nvidia.com/vgpu/deployment/vmware/latest/index

Installs the NVIDIA vGPU driver on a Linux virtual machine, specifically tested with Ubuntu. This involves using package managers or direct installation scripts. Dependencies such as kernel headers may need to be installed prior to driver installation.

```text
1. Ensure your Ubuntu VM has the necessary build tools and kernel headers installed:
   sudo apt update
   sudo apt install build-essential linux-headers-$(uname -r)

2. Download the NVIDIA vGPU driver for Linux from the NVIDIA Enterprise Support website.
3. Make the downloaded driver script executable:
   chmod +x NVIDIA-Linux-x86_64-XXX.XX.run

4. Run the installer with root privileges:
   sudo ./NVIDIA-Linux-x86_64-XXX.XX.run

5. Follow the on-screen prompts. It is recommended to let the installer update your xorg.conf file if prompted.
6. Reboot the virtual machine after the installation is complete.
```

--------------------------------

### Check NVIDIA Control Panel Installation using PowerShell

Source: https://docs.nvidia.com/vgpu/17.0/grid-vgpu-release-notes-generic-linux-kvm/index

This snippet demonstrates how to use the Get-AppxPackage PowerShell command to determine if the NVIDIA Control Panel is installed on a Windows system. It shows commands for checking the current user and all users, and an example of the output indicating installation status.

```powershell
Get-AppxPackage -Name NVIDIACorp.NVIDIAControlPanel
```

```powershell
Get-AppxPackage -AllUsers -Name NVIDIACorp.NVIDIAControlPanel
```

```powershell
Get-AppxPackage -AllUsers -Name NVIDIACorp.NVIDIAControlPanel

Name                   : NVIDIACorp.NVIDIAControlPanel
Publisher              : CN=D6816951-877F-493B-B4EE-41AB9419C326
Architecture           : X64
ResourceId             :
Version                : 8.1.964.0
PackageFullName        : NVIDIACorp.NVIDIAControlPanel_8.1.964.0_x64__56jybvy8sckqj
InstallLocation        : C:\Program Files\WindowsApps\NVIDIACorp.NVIDIAControlPanel_8.1.964.0_x64__56jybvy8sckqj
IsFramework            : False
PackageFamilyName      : NVIDIACorp.NVIDIAControlPanel_56jybvy8sckqj
PublisherId            : 56jybvy8sckqj
PackageUserInformation : {S-1-12-1-530092550-1307989247-1105462437-500 **[Administrator]: Installed**,
                         S-1-12-1-530092550-1307989247-1105462437-1002 **[pliny]: Installed**,
                         S-1-12-1-530092550-1307989247-1105462437-1003 **[trajan]: Installed**}
IsResourcePackage      : False
IsBundle               : False
IsDevelopmentMode      : False
NonRemovable           : False
IsPartiallyStaged      : False
SignatureKind          : Store
Status                 : Ok
```

--------------------------------

### Start x11vnc server on display 0

Source: https://docs.nvidia.com/vgpu/10.0/grid-vgpu-user-guide

Starts the x11vnc server, connecting to the existing X session on display :0. It includes options for authentication, sharing, and background operation.

```bash
# x11vnc -display :0 -auth /run/user/121/gdm/Xauthority -forever \
-shared -ncache -bg
```

--------------------------------

### Run NVIDIA vGPU Driver Installer with Kernel Source Path

Source: https://docs.nvidia.com/vgpu/10.0/grid-vgpu-user-guide

This command runs the NVIDIA vGPU software Linux driver installer and explicitly specifies the path to the kernel source. This is used when the installer fails to automatically detect installed kernel headers and sources. The path provided is an example and should be adjusted based on the system's kernel version.

```bash
sudo sh ./ NVIDIA-Linux_x86_64-352.47-grid.run \
–kernel-source-path=/usr/src/kernels/3.10.0-229.11.1.el7.x86_64
```

--------------------------------

### Start Xorg Server

Source: https://docs.nvidia.com/vgpu/11.0/grid-vgpu-release-notes-microsoft-windows-server/index

This command is used to start the Xorg server. It's the counterpart to stopping the server and is necessary for graphical environments to function.

```shell
# etc/init.d/xorg start

```

--------------------------------

### Example gridd.conf for NVIDIA vGPU

Source: https://docs.nvidia.com/vgpu/14.0/grid-licensing-user-guide/index

Provides an example configuration file for NVIDIA vGPU licensing, demonstrating how to set the primary and backup license server addresses and ports, and enabling vGPU by setting FeatureType to 1.

```ini
# /etc/nvidia/gridd.conf.template - Configuration file for NVIDIA Grid Daemon

# This is a template for the configuration file for NVIDIA Grid Daemon.
# For details on the file format, please refer to the nvidia-gridd(1)
# man page.

# Description: Set License Server Address
# Data type: string
# Format:  "<address>"
ServerAddress= gridlicense1.example.com

# Description: Set License Server port number
# Data type: integer
# Format:  <port>, default is 7070
ServerPort=7070

# Description: Set Backup License Server Address
# Data type: string
# Format:  "<address>"
BackupServerAddress= gridlicense2.example.com

# Description: Set Backup License Server port number
# Data type: integer
# Format:  <port>, default is 7070
BackupServerPort=7070

# Description: Set Feature to be enabled
# Data type: integer
# Possible values:
#    0 => for unlicensed state
#    1 => for GRID vGPU
#    2 => for Quadro Virtual Datacenter Workstation
#    4 => for NVIDIA Virtual Compute Server
FeatureType=1

# Description: Parameter to enable or disable Grid Licensing tab in nvidia-settings
# Data type: boolean
# Possible values: TRUE or FALSE, default is FALSE
#EnableUI=TRUE

# Description: Set license borrow period in minutes
# Data type: integer
# Possible values: 10 to 10080 mins(7 days), default is 1440 mins(1 day)
#LicenseInterval=1440

# Description: Set license linger period in minutes
# Data type: integer
# Possible values: 0 to 10080 mins(7 days), default is 0 mins
#LingerInterval=10
```

--------------------------------

### Install NVIDIA vGPU Driver on Windows VM

Source: https://docs.nvidia.com/vgpu/deployment/vmware/latest/index

Installs the NVIDIA vGPU driver on a Microsoft Windows virtual machine. This process typically involves downloading the driver installer and running it within the VM. Ensure the correct driver version is selected based on the vGPU profile and host configuration.

```text
1. Download the appropriate NVIDIA vGPU driver for Windows from the NVIDIA Enterprise Support website.
2. Extract the downloaded driver package.
3. Run the driver installer executable (e.g., setup.exe).
4. Follow the on-screen prompts to complete the installation.
5. Reboot the virtual machine after the installation is complete.
```

--------------------------------

### Example gridd.conf for Quadro vDWS

Source: https://docs.nvidia.com/vgpu/9.0/grid-licensing-user-guide/index

An example of a gridd.conf file specifically configured for Quadro vDWS. This configuration sets the primary and backup license servers, their ports, and the FeatureType to 2.

```ini
# /etc/nvidia/gridd.conf.template - Configuration file for NVIDIA Grid Daemon

# This is a template for the configuration file for NVIDIA Grid Daemon.
# For details on the file format, please refer to the nvidia-gridd(1)
# man page.

# Description: Set License Server Address
ServerAddress=gridlicense1.example.com
BackupServerAddress=gridlicense2.example.com
ServerPort=7070
BackupServerPort=7070
FeatureType=2
```

--------------------------------

### Example of RPM Package Conflict Errors during Update

Source: https://docs.nvidia.com/vgpu/4.6/grid-vgpu-user-guide/index

Illustrates potential conflict errors encountered when attempting to install a newer NVIDIA vGPU Manager RPM package without using the upgrade option ('-U'). This indicates a version mismatch and prevents installation.

```shell
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-367.124.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.123.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-367.124.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.123.x86_64
        ...

```

--------------------------------

### Building NVML-enabled Applications for vGPU Hosts

Source: https://docs.nvidia.com/vgpu/4.7/grid-management-sdk-user-guide/index

Guidance on including necessary headers to build NVML-enabled applications for vGPU hosts.

```APIDOC
## Building NVML-enabled Applications for vGPU Hosts

### Description
Information on the header files required to build applications that utilize NVML for vGPU hosts.

### Usage

To build an NVML-enabled application for a vGPU host, ensure that you include `nvml_grid.h` in addition to `nvml.h`.

### Code Example

```c
#include <nvml.h>
#include <nvml_grid.h>
```

For more information, refer to the sample code included in the SDK.
```

--------------------------------

### Building NVML-enabled Applications for vGPU Hosts

Source: https://docs.nvidia.com/vgpu/12.0/grid-management-sdk-user-guide/index

Instructions on how to include the necessary headers for building NVML-enabled applications on vGPU hosts.

```APIDOC
## Building NVML-enabled Applications for vGPU Hosts

### Description
To build an NVML-enabled application for a vGPU host, you need to include specific header files.

### Header Files
Ensure that you include `nvml_grid.h` in addition to the standard `nvml.h`.

### Code Example
```c
#include <nvml.h>
#include <nvml_grid.h>
```

For more detailed information, refer to the sample code provided in the SDK.
```

--------------------------------

### Building NVML-enabled Application for vGPU Host

Source: https://docs.nvidia.com/vgpu/4.6/grid-management-sdk-user-guide/index

Instructions on how to build an NVML-enabled application for a vGPU host by including the necessary header files.

```APIDOC
## Building an NVML-enabled Application for a vGPU Host

### Description
Instructions on building an NVML-enabled application for a vGPU host.

### Requirements
- Include `nvml_grid.h` in addition to `nvml.h`.

### Code Example
```c
#include <nvml.h>
#include <nvml_grid.h>
```

For more information, refer to the sample code included in the SDK.
```

--------------------------------

### Example of NVIDIA GPU Manager Package Conflict Errors

Source: https://docs.nvidia.com/vgpu/10.0/grid-vgpu-user-guide

This output demonstrates potential conflict errors when installing a newer NVIDIA GRID package without using the upgrade option (`-U`). The `rpm` command reports that files from the new package conflict with existing files from a previously installed version, indicating an improper installation or upgrade process. This is relevant for hypervisors using RPM package management.

```shell
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-440.121.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-440.107.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-440.121.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-440.107.x86_64
        ...

        

```

--------------------------------

### Example of NVIDIA GPU Manager Package Conflict on XenServer

Source: https://docs.nvidia.com/vgpu/4.6/grid-vgpu-user-guide/index

This output demonstrates conflict errors when attempting to install a newer NVIDIA GPU Manager package without using the upgrade option on Citrix XenServer. It indicates that existing files from a previous version conflict with the new installation, requiring a clean upgrade.

```bash
[root@xenserver ~]# rpm –q NVIDIA-vGPU-xenserver-7.0-367.124
[root@xenserver ~]#
If an existing NVIDIA GRID package is already installed and you don’t select the upgrade (-U) option when installing a newer GRID package, the rpm command will return many conflict errors.
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-367.124.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.123.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-367.124.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.123.x86_64
        ...

```

--------------------------------

### Building NVML-enabled Applications for vGPU Hosts

Source: https://docs.nvidia.com/vgpu/4.5/grid-management-sdk-user-guide/index

Information on how to build an NVML-enabled application for a vGPU host, including the necessary header files.

```APIDOC
## Building an NVML-enabled application for a vGPU host

### Description
Guidance on compiling applications that utilize NVIDIA Management Library (NVML) for vGPU hosts.

### Header Files
Functions specific to vGPUs are defined in the `nvml_grid.h` header file. Ensure you include both `nvml.h` and `nvml_grid.h`.

### Code Example
```c
#include <nvml.h>
#include <nvml_grid.h>
```

For more information, refer to the sample code included in the SDK.
```

--------------------------------

### Get Xorg server display numbers

Source: https://docs.nvidia.com/vgpu/10.0/grid-vgpu-user-guide

Retrieves the display numbers for active Xorg server sessions by parsing environment variables. This is a prerequisite for starting the x11vnc server.

```bash
# cat /proc/*/environ 2>/dev/null | tr '\0' '\n' | grep '^DISPLAY=:' | uniq
```

--------------------------------

### Start nvidia-gridd Service

Source: https://docs.nvidia.com/vgpu/10.0/grid-licensing-user-guide/index

Command to start the nvidia-gridd service, typically required after configuring license management settings. This command requires sudo privileges.

```bash
# sudo service nvidia-gridd start

```

--------------------------------

### Start and Stop Xorg Server

Source: https://docs.nvidia.com/vgpu/10.0/grid-vgpu-release-notes-citrix-xenserver/index

These commands are used to control the Xorg display server. The 'stop' command halts the server, while the 'start' command initiates it. These are fundamental operations for managing graphical environments on Linux-based systems.

```bash
# /etc/init.d/xorg stop
```

```bash
# etc/init.d/xorg start
```

--------------------------------

### Example Warnings During vGPU Manager Installation/Uninstallation

Source: https://docs.nvidia.com/vgpu/19.0/known-issues/bug-200423757-vgpu-manager-installation-benign-messages

This snippet shows example warning messages related to missing kernel modules and files that may appear during the installation or uninstallation of the Virtual GPU Manager on a system after a hypervisor upgrade. These warnings are benign and do not indicate a failure in the process.

```shell
WARNING: Can't read module /lib/modules/4.4.77-1.el6.nutanix.20170830.124.x86_64/weak-updates/ixgbevf.ko: No such file or directory
WARNING: Can't read module /lib/modules/4.4.77-1.el6.nutanix.20170830.124.x86_64/weak-updates/i40evf.ko: No such file or directory
egrep: /lib/modules/4.4.77-1.el6.nutanix.20170830.124.x86_64//weak-updates/ixgbevf.ko: No such file or directory
egrep: /lib/modules/4.4.77-1.el6.nutanix.20170830.124.x86_64//weak-updates/i40evf.ko: No such file or directory
gzip: /boot/initramfs-4.4.77-1.el6.nutanix.20170830.124.x86_64.tmp: not in gzip format
WARNING: Can't read module /lib/modules/4.4.77-1.el6.nutanix.20170830.124.x86_64/weak-updates/ixgbevf.ko: No such file or directory
WARNING: Can't read module /lib/modules/4.4.77-1.el6.nutanix.20170830.124.x86_64/weak-updates/i40evf.ko: No such file or directory
```

--------------------------------

### Example Uncorrectable Machine Check Exception Log

Source: https://docs.nvidia.com/vgpu/10.0/grid-vgpu-release-notes-red-hat-el-kvm/index

This log snippet shows example error messages associated with an uncorrectable machine check exception occurring during the POST phase after installing the Virtual GPU Manager on HP systems running newer Linux kernels. These errors indicate issues with the PCI Express interface.

```log
Uncorrectable Machine Check Exception (Processor 1, APIC ID 0x00000000, Bank 0x00000006, 
Status 0xFB800000'00000E0B, Address 0x00000000'00000000, Misc 0x00000000'5B000000).
Uncorrectable PCI Express Error Detected. Slot 255 (Segment 0x0, Bus 0x5D, Device 0x48, Function 0x0)
Uncorrectable PCI Express Error Detected. Slot 255 (Segment 0x0, Bus 0x5D, Device 0x88, Function 0x0)
Uncorrectable PCI Express Error Detected. Slot 255 (Segment 0x0, Bus 0x5D, Device 0x80, Function 0x0)
Uncorrectable PCI Express Error Detected. Slot 255 (Segment 0x0, Bus 0x5D, Device 0x40, Function 0x0)
```

--------------------------------

### Create vGPU Instance (Bash)

Source: https://docs.nvidia.com/vgpu/latest/grid-vgpu-user-guide

This command creates an instance of a vGPU by writing the UUID to a 'create' file within the specified subdirectory. The subdirectory contains registration information for the vGPU type. This is a fundamental step in provisioning a vGPU.

```bash
# echo "_uuid_"> _subdirectory_/create

_uuid_ 
     The UUID that you generated in the previous step, which will become the UUID of the vGPU that you want to create.  

_subdirectory_ 
     The registration information directory for the vGPU type that you want to create, for example, `nvidia-41`. 
This example creates an instance of the M10-2Q vGPU type with the UUID `aa618089-8b16-4d01-a136-25a0f3c73123`.
# echo "aa618089-8b16-4d01-a136-25a0f3c73123" > nvidia-41/create
```

--------------------------------

### RPM Package Conflict Example

Source: https://docs.nvidia.com/vgpu/latest/grid-vgpu-user-guide

Illustrates potential conflict errors when installing a newer NVIDIA GRID package without the upgrade option (`-U`), showing conflicts with existing files.

```shell
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-NVIDIA-vGPU-CitrixHypervisor-8.2-580.95.02.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-8.2-580.65.05.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-NVIDIA-vGPU-CitrixHypervisor-8.2-580.95.02.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-8.2-580.65.05.x86_64
        ...

```

--------------------------------

### Build NVML Application for Guest VM - C++

Source: https://docs.nvidia.com/vgpu/18.0/grid-management-sdk-user-guide/index

Instructions on building an NVML-enabled application for a guest VM. This involves setting up the build environment and compiling code that utilizes the NVML library to interact with the GPU within the virtual machine.

```c++
#include <nvml.h>
#include <iostream>

int main() {
    nvmlInit();
    // NVML application code here
    nvmlShutdown();
    return 0;
}
```

--------------------------------

### Building NVML Application for vGPU Host

Source: https://docs.nvidia.com/vgpu/4.2/grid-management-sdk-user-guide/index

Instructions for building an NVML-enabled application for a vGPU host.

```APIDOC
## Building an NVML-enabled application for a vGPU host

Functions that are specific to vGPUs are defined in the header file `nvml_grid.h`.

To build an NVML-enabled application for a vGPU host, ensure that you include `nvml_grid.h` in addition to `nvml.h`.

```c
#include <nvml.h>
#include <nvml_grid.h>
```

For more information, refer to the sample code that is included in the SDK.
```

--------------------------------

### Get vGPU Serial Number Property (Not Supported)

Source: https://docs.nvidia.com/vgpu/12.0/grid-management-sdk-user-guide/index

This example shows the NVML function nvmlDeviceGetSerial() and its expected return code when querying a vGPU. For vGPUs, serial numbers are not assigned, and this function will return NVML_ERROR_NOT_SUPPORTED.

```c
#include <nvml.h>

// ... (NVML initialization and device handle obtained) ...

char serialNumber[256];
nvmlReturn_t result;

result = nvmlDeviceGetSerial(device, serialNumber, sizeof(serialNumber));
// For a vGPU, result is expected to be NVML_ERROR_NOT_SUPPORTED
// The serialNumber buffer will contain an undefined value.
```

--------------------------------

### Check Running Xorg Server Process (Ubuntu Example)

Source: https://docs.nvidia.com/vgpu/11.0/grid-vgpu-user-guide

This command checks if the Xorg server process is currently running on an Ubuntu system. It is used to confirm that the Xorg server has started successfully after configuration changes.

```shell
# ps -ef | grep X

root     16500 16499  2 03:01 tty2     00:00:00 /usr/lib/xorg/Xorg -nolisten tcp :0 -auth /tmp/serverauth.s7CE4mMeIz
root      1140  1126  0 18:46 tty1     00:00:00 /usr/lib/xorg/Xorg vt1 -displayfd 3 -auth /run/user/121/gdm/Xauthority -background none -noreset -keeptty -verbose 3
root     17011 17108  0 18:50 pts/0    00:00:00 grep --color=auto X
```

--------------------------------

### Blacklist Nouveau Driver Configuration

Source: https://docs.nvidia.com/vgpu/toolkits/fine-tunning-llms/latest/deployment

This snippet shows how to configure the system to blacklist the nouveau driver and disable its modesetting. This is a crucial step before installing NVIDIA's proprietary drivers to avoid conflicts.

```bash
$ sudo vim /etc/modprobe.d/blacklist.conf
$ blacklist nouveau
$ options nouveau modeset=0
```

--------------------------------

### Example gridd.conf for GRID vGPU (NVIDIA Configuration)

Source: https://docs.nvidia.com/vgpu/4.7/grid-licensing-user-guide/index

An example configuration file for the NVIDIA Grid Daemon, specifying the license server address, port, and feature type for GRID vGPU. This file is used to enable vGPU licensing.

```ini
# /etc/nvidia/gridd.conf - Configuration file for NVIDIA Grid Daemon
        
# This is a template for the configuration file for NVIDIA Grid Daemon.
# For details on the file format, please refer to the nvidia-gridd(1)
# man page.

# Description: Set License Server Address
# Data type: string
# Format:  "<address>"
ServerAddress=gridlicense.example.com

# Description: Set License Server port number
# Data type: integer
# Format: <port>, default is 7070
ServerPort=7070

# Description: Set Feature to be enabled
# Data type: integer
# Possible values:
#    1 => for GRID vGPU
#    2 => for GRID Virtual Workstation
FeatureType=1

# Description: Parameter to enable or disable Grid Licensing tab in nvidia-settings
# Data type: boolean
# Possible values: TRUE or FALSE, default is TRUE
#EnableUI=TRUE

# Description: Set license borrow period in minutes
# Data type: integer
# Possible values: 10 to 10080 mins(7 days), default is 10080
#Licenselnterval=10080

        
```

--------------------------------

### Blacklist Nouveau Driver

Source: https://docs.nvidia.com/vgpu/toolkits/rag-application/latest/deployment

This snippet demonstrates how to blacklist the nouveau driver and disable its modesetting. This is a crucial step before installing NVIDIA drivers. It involves editing a configuration file and adding specific lines.

```bash
sudo vim /etc/modprobe.d/blacklist.conf
blacklist nouveau
options nouveau modeset=0
```

--------------------------------

### Example gridd.conf File for NVIDIA vGPU

Source: https://docs.nvidia.com/vgpu/5.0/grid-licensing-user-guide/index

An example configuration file for NVIDIA vGPU licensing on Linux. It demonstrates how to set the primary and backup license server addresses and ports, and enables vGPU licensing by setting FeatureType to 1.

```ini
# /etc/nvidia/gridd.conf.template - Configuration file for NVIDIA Grid Daemon

# This is a template for the configuration file for NVIDIA Grid Daemon.
# For details on the file format, please refer to the nvidia-gridd(1)
# man page.

# Description: Set License Server Address
# Data type: string
# Format:  "<address>"
ServerAddress= gridlicense1.example.com

# Description: Set License Server port number
# Data type: integer
# Format:  <port>, default is 7070
ServerPort=7070

# Description: Set Backup License Server Address
# Data type: string
# Format:  "<address>"
BackupServerAddress= gridlicense2.example.com

# Description: Set Backup License Server port number
# Data type: integer
# Format:  <port>, default is 7070
BackupServerPort=7070

# Description: Set Feature to be enabled
# Data type: integer
# Possible values:
#    0 => for unlicensed state
#    1 => for GRID vGPU
#    2 => for Quadro Virtual Datacenter Workstation
FeatureType=1

# Description: Parameter to enable or disable Grid Licensing tab in nvidia-settings
# Data type: boolean
# Possible values: TRUE or FALSE, default is FALSE
#EnableUI=TRUE

# Description: Set license borrow period in minutes
# Data type: integer
# Possible values: 10 to 10080 mins(7 days), default is 1440 mins(1 day)
#LicenseInterval=1440

# Description: Set license linger period in minutes
# Data type: integer
# Possible values: 0 to 10080 mins(7 days), default is 0 mins
#LingerInterval=10

```

--------------------------------

### Check Running Xorg Server Process (RHEL Example)

Source: https://docs.nvidia.com/vgpu/11.0/grid-vgpu-user-guide

This command checks if the Xorg server process is currently running on a Red Hat Enterprise Linux system. It is used to confirm that the Xorg server has started successfully after configuration changes.

```shell
# ps -ef | grep X

root      5285  5181  0 16:29 pts/0    00:00:00 grep --color=auto X
root      5880     1  0 Jun13 ?        00:00:00 /usr/bin/abrt-watch-log -F Backtrace /var/log/Xorg.0.log -- /usr/bin/abrt-dump-xorg -xD
root      7039  6289  0 Jun13 tty1     00:00:03 /usr/bin/X :0 -background none -noreset -audit 4 -verbose -auth /run/gdm/auth-for-gdm-vr4MFC/database -seat seat0 vt1
```

--------------------------------

### Example gridd.conf for GRID Virtual Workstation

Source: https://docs.nvidia.com/vgpu/4.10/grid-licensing-user-guide/index

This is an example of the gridd.conf file used to configure NVIDIA GRID Daemon for GRID Virtual Workstation. It specifies the license server address, port, and the feature type to enable.

```ini
# /etc/nvidia/gridd.conf - Configuration file for NVIDIA Grid Daemon
        
# This is a template for the configuration file for NVIDIA Grid Daemon.
# For details on the file format, please refer to the nvidia-gridd(1)
# man page.

# Description: Set License Server Address
# Data type: string
# Format:  "<address>"
ServerAddress=gridlicense.example.com

# Description: Set License Server port number
# Data type: integer
# Format: <port>, default is 7070
ServerPort=7070

# Description: Set Feature to be enabled
# Data type: integer
# Possible values:
#    1 => for GRID vGPU
#    2 => for GRID Virtual Workstation
```

--------------------------------

### Example gridd.conf File for NVIDIA vGPU

Source: https://docs.nvidia.com/vgpu/12.0/grid-licensing-user-guide/index

This is a template for the '/etc/nvidia/gridd.conf' configuration file used by the NVIDIA Grid Daemon on Linux systems. It demonstrates how to set parameters like server addresses, ports, and the feature type for licensing NVIDIA vGPU software.

```ini
# /etc/nvidia/gridd.conf.template - Configuration file for NVIDIA Grid Daemon

# This is a template for the configuration file for NVIDIA Grid Daemon.
```

--------------------------------

### Verify NVIDIA GPU Manager Package on XenServer

Source: https://docs.nvidia.com/vgpu/4.10/grid-vgpu-user-guide/index

This command verifies the installation of the NVIDIA GPU Manager RPM package on Citrix XenServer. It's essential for ensuring that the necessary management components for vGPU are present. The example shows how to query a specific package version.

```shell
rpm -q _vgpu-manager-rpm-package-name_

[root@xenserver ~]# rpm –q NVIDIA-vGPU-xenserver-7.0-367.134
[root@xenserver ~]#
If an existing NVIDIA GRID package is already installed and you don’t select the upgrade (-U) option when installing a newer GRID package, the rpm command will return many conflict errors.
Preparing packages for installation...
        file /usr/bin/nvidia-smi from install of NVIDIA-vGPU-xenserver-7.0-367.134.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.132.x86_64
        file /usr/lib/libnvidia-ml.so from install of NVIDIA-vGPU-xenserver-7.0-367.134.x86_64 conflicts with file from package NVIDIA-vGPU-xenserver-7.0-367.132.x86_64
        ...


```

--------------------------------

### Get GPU BAR1 Memory Usage with nvidia-smi

Source: https://docs.nvidia.com/vgpu/latest/grid-vgpu-release-notes-vmware-vsphere/index

This command retrieves detailed information about NVIDIA GPUs, including the BAR1 memory usage which is critical for calculating MMIO space requirements when assigning GPUs in pass-through mode. Ensure the NVIDIA driver is installed on the hypervisor host.

```bash
nvidia-smi -q
==============NVSMI LOG==============

Timestamp                                 : Mon Oct 13 18:36:45 2025
Driver Version                            : 580.95.02 
CUDA Version                              :  13.0

Attached GPUs                             : 4
GPU 00000000:01:00.0
...
   **BAR1 Memory Usage Total : 128 GiB**
...
        
```

--------------------------------

### Example gridd.conf for vWS Licensing (Linux)

Source: https://docs.nvidia.com/vgpu/12.0/grid-licensing-user-guide/index

This is an example of a gridd.conf file configured for vWS licensing. It specifies the primary and backup license server addresses and ports, and sets the FeatureType to 2 for vWS.

```ini
# /etc/nvidia/gridd.conf.template - Configuration file for NVIDIA Grid Daemon

# Description: Set License Server Address
# Data type: string
# Format:  "<address>"
ServerAddress= gridlicense1.example.com

# Description: Set License Server port number
# Data type: integer
# Format:  <port>, default is 7070
ServerPort=7070

# Description: Set Backup License Server Address
# Data type: string
# Format:  "<address>"
BackupServerAddress= gridlicense2.example.com

# Description: Set Backup License Server port number
# Data type: integer
# Format:  <port>, default is 7070
BackupServerPort=7070

# Description: Set Feature to be enabled
# Data type: integer
# Possible values:
vWS | 2
vCS | 4
FeatureType= 2
```
Tokens:
Raw