SageMaker JupyterLab Ray Extension

A JupyterLab 4 extension for discovering and connecting to Ray clusters running on SageMaker HyperPod (EKS-orchestrated) from within SageMaker Studio.

Overview

This extension adds a Ray Clusters tab to the "Connect to cluster" dialog in the JupyterLab notebook toolbar. Users can:

• Browse RayClusters across every namespace they can access (paginated, with a per-namespace fallback when cluster-wide listing is not permitted)
• View cluster status, Ray version, workers, GPUs, vCPUs, and namespace
• Insert a job submission code snippet into the active notebook cell
• Deep-link to cluster details in the HyperPod console

Data Layer

Ray cluster data is fetched via Skylens — AWS's managed Kubernetes API proxy (eks-proxy.eks.<region>.amazonaws.com), authenticated with SigV4 using the space execution role. No kubectl or kubeconfig required.

Note: The HyperPod default access policies do not grant ray.io API access by default. A cluster admin must apply a Kubernetes RBAC Role+RoleBinding granting ray.io read access to the space's execution role.

Development

brazil-build release

• This package on its own (build, unit tests, lint, watch mode, the jp_fetch test note): see DEVELOPING.md.
• Testing Ray together with the Common and EMR extensions (the shared "Connect to cluster" dialog, end to end): see ../SageMakerStudioJupyterLabExtensionCommon/E2E_TESTING.md.