KubeRay

Install Ray on Kubernetes.

Ray is a project for running machine learning jobs in a cluster of machines. KubeRay is an operator for installing Ray on top of Kubernetes. For up-to-date installation guide check with the upstream Ray documentation.

Install Ray operator with Helm

Create a Kubernetes cluster with talosctl or via Omni.

Deploy the operator via Helm.

helm repo add kuberay https://ray-project.github.io/kuberay-helm/
helm repo update

helm install kuberay-operator kuberay/kuberay-operator --version 1.4.2

Once the operator pod is running you can create a Ray cluster.

helm install raycluster kuberay/ray-cluster --version 1.4.2

Get the status of the cluster:

kubectl get rayclusters

The output should look like this.

NAME                 DESIRED WORKERS   AVAILABLE WORKERS   CPUS   MEMORY   GPUS   STATUS   AGE
raycluster-kuberay   1                 1                   2      3G       0      ready    13m

From here you can follow the Ray documentation to run a RayJob.