Introducing managed cluster extensions for observability and user management in Amazon SageMaker HyperPod Slurm
Introducing managed cluster extensions for observability and user management in Amazon SageMaker HyperPod Slurm
Amazon SageMaker HyperPod launched cluster extensions, AWS-managed software packages installed, configured, updated, and removed through APIs or the SageMaker AI console, starting with the sagemaker-observability extension (node, GPU, EFA, and Slurm metrics to CloudWatch or Amazon Managed Service for Prometheus) and the sagemaker-user-management extension (consistent users and groups across nodes, SSH and sudo rules, Slurm accounting, and Active Directory integration). HyperPod validates extension settings against a schema before rollout, applies them to every node including new and replaced ones, and reports per-node status and health, while users retain full node access including the controller. The post also covers the declarative features introduced earlier in 2026 and a four-step path for migrating capabilities from lifecycle scripts to extensions.
Source: AWS Machine Learning Blog · aws.amazon.com