Installing KubeRay on HyperPod Amazon EKS
The KubeRay operator manages the lifecycle of RayCluster,
RayJob, RayCronJob, and RayService resources. It is
required for every Ray capability on HyperPod. HyperPod uses the open
source operator without any changes.
Prerequisites
-
A HyperPod cluster orchestrated by Amazon EKS. For more information, see Creating a SageMaker HyperPod cluster with Amazon EKS orchestration.
-
kubectlandhelminstalled, withkubectlconfigured for your cluster:aws eks update-kubeconfig --namemy-eks-cluster--regionmy-region
Installing on an existing cluster
If your cluster does not already run KubeRay, install the operator with Helm.
To install the KubeRay operator
-
Add the KubeRay Helm repository.
helm repo add kuberay https://ray-project.github.io/kuberay-helm/ helm repo update -
Install the operator.
helm install kuberay-operator kuberay/kuberay-operator -
Confirm that the operator pod is running.
kubectl get pods -l app.kubernetes.io/name=kuberay-operator
For more information about operator installation options, see KubeRay operator installation
Upgrading and uninstalling
Upgrade the operator to the latest chart version:
helm repo update helm upgrade kuberay-operator kuberay/kuberay-operator
Uninstalling removes the operator but not your Ray custom resources:
helm uninstall kuberay-operator
Warning
Uninstalling the operator stops reconciliation of every Ray resource in the cluster. Running Ray clusters continue but are no longer managed, and deletions do not complete. Delete your Ray resources before you uninstall.
