Deploy HyperPod EKS infrastructure using deploy.sh via CloudFormation or Terraform, including AZ resolution, stack outputs, and kubeconfig setup
Use this skill to deploy the HyperPod EKS infrastructure that hosts the Slurm
cluster. This is Phase 1 of the slinky-slurm deployment workflow and must
complete before running setup.sh or install.sh.
The deploy.sh script supports two infrastructure backends:
--infra cfn) -- uses an AWS-hosted S3 template--infra tf) -- uses local Terraform modulesDeployment takes approximately 20-30 minutes and provisions:
Before running deploy.sh, verify:
--infra cfn) or terraform installed (for
--infra tf)hp-eks-slinky-stack)See the deployment-preflight skill for detailed prerequisite validation.
Any SageMaker-supported GPU instance type can be used. GPU count, model, and EFA interfaces are auto-discovered via the EC2 API. Two common