NVIDIA GPU Operator

Kubernetes provides access to special hardware resources such as NVIDIA GPUs, NICs, Infiniband adapters and other devices through the device plugin framework. However, configuring and managing nodes with these hardware resources requires configuration of multiple software components such as drivers, container runtimes or other libraries which are difficult and prone to errors. The NVIDIA GPU Operator uses the operator framework within Kubernetes to automate the management of all NVIDIA software components needed to provision GPU. These components include the NVIDIA drivers (to enable CUDA), Kubernetes device plugin for GPUs, the NVIDIA Container Toolkit, automatic node labelling using GFD, DCGM based monitoring and others. To know more, click here.

Prerequisite

Ensure that the input/software_config.json file contains the following line under softwares:

{"name": "nvidia_gpu_operator", "version":"25.3.0"}

A sample version of the input/software_config.json is located here.

Note

Currently, Omnia only supports the 25.3.0 version of the NVIDIA GPU operator.

Configurations

Omnia installs the NVIDIA GPU operator as part of omnia.yml or scheduler.yml playbook execution, based on if the input/omnia_config.yml file contains the required inputs. Follow the below provided steps to set up the NVIDIA GPU operator for your cluster:

  1. Edit the nvidia_gpu_operator_config.yml file present in the omnia/input/ folder. Provide inputs to the following mandatory parameters in that file:

    • http_proxy: This value can be found in the /opt/omnia/offline/local_repo_access.yml file.

    • https_proxy: This value can be found in the /opt/omnia/offline/local_repo_access.yml file.

Note

Omnia supports the successful deployment of the NVIDIA GPU Operator for the pre-filled basic configuration, under input/nvidia_gpu_operator. For a more customized installation, the configuration file can be edited as per the user’s requirements. However, users must check and verify the compatibility of supported tools and driver versions before proceeding. Omnia does not claim responsibility for any issues arising from custom modifications or compatibility checks.

  1. Once the nvidia_gpu_operator_config.yml file is ready, by default the filepath is added to the nvidia_gpu_operator_value_file_path parameter in input/omnia_config.yml file.

Playbook execution

Once the above mentioned configurations are done, execute the omnia.yml or scheduler.yml playbook to install the NVIDIA GPU operator:

cd omnia
ansible-playbook omnia.yml -i <inventory filepath>

cd scheduler
ansible-playbook scheduler.yml -i <inventory filepath>

Here, the <inventory filepath> refers to the Kubernetes inventory.

Remove the NVIDIA GPU operator

If you want to uninstall the NVIDIA GPU operator, run the following command:

helm delete gpu-operator -n gpu-operator

If you have any feedback about Omnia documentation, please reach out at omnia.readme@dell.com.