GitHub - NVIDIA/k8s-nim-operator: An Operator for deployment and maintenance of NVIDIA NIMs and NeMo microservices in a Kubernetes environment. (original) (raw)

NVIDIA NIM Operator

NVIDIA NIM Operator is a Kubernetes Operator that is designed to facilitate the deployment, management, and scaling of NVIDIA NIM microservices on Kubernetes clusters.

NVIDIA NIM microservices deliver AI foundation models as accelerated inference microservices that are portable across data center, workstation, and cloud, accelerating flexible generative AI development, deployment and time to value.

To use the Operator in your cluster, refer to docsfor installation and configuration information.

Developing the Operator

Prerequisites

Kubernetes v1.28 and higher.
NVIDIA GPUs that are supported by the NIM microservices to deploy.

Deploying the Operator on the Cluster

Build and push your image to the location specified by IMAGE_NAME:VERSION:

make build IMAGE_NAME=/k8s-nim-operator VERSION= -f deployments/container/Makefile

Publish the image to a personal registry. You must be able to pull the image from the working environment. Make sure you have the proper permission to the registry if the preceding commands result in an error.

Install the CRDs into the cluster:

Deploy the manager to the cluster with the image specified by IMG:

make deploy IMG=/k8s-nim-operator:

If you encounter RBAC errors, you might need to grant yourself cluster-admin privileges or be logged in as admin. Ensure that the samples have default values.

Deploying Sample `NIMCache`, `NIMService`, `NIMPipeline`, `NemoCustomizer`, `NemoEvaluator` and `NemoGuardrail` Resources

Follow the guides in the docs to deploy sample CR instances.

Uninstalling the Operator

Delete the instances (CRs) from the cluster:

kubectl delete -k config/samples/

Delete the APIs (CRDs) from the cluster:

Undeploy the controller from the cluster:

Project Distribution

Perform the following steps to build the installation manifests and distribute this project to users.

Build the manifests for the image built and published in the registry:
make build-installer IMG=/k8s-nim-operator:
The preceding Makefile target generates a dist/install.yaml file. This file is built with Kustomize and contains the manifests for the CRDs and resources that are necessary to install this project without its dependencies.
Run the installer:
kubectl apply -f https://raw.githubusercontent.com//k8s-nim-operator//dist/install.yaml

Contributing

NVIDIA can work with partners to add platform support for the NIM Operator. The NIM Operator is open-source and permissively licensed with only minimal requirements for source code contributions.

To get started with building the NIM Operator, follow these steps:

git clone git@github.com:NVIDIA/k8s-nim-operator.git cd k8s-nim-operator make build IMAGE_NAME=/k8s-nim-operator VERSION= -f deployments/container/Makefile

Run make help for more information about additional make targets.

More information can be found in the Kubebuilder Documentation.