Cloud Cluster Guide

FAQ | Troubleshooting | Glossary

• Cloud Cluster Guide
  • Overview
  • Cluster Toolkit
  • Terraform
  • Quickstart Examples

Overview

This guide focuses on setting up a cloud Slurm cluster. Deploying in a cloud environment involves various decisions and considerations, which this guide will cover, along with recommended solutions.

The primary deployment method for cloud cluster management is:

• Cluster Toolkit

Cluster Toolkit

The Cluster Toolkit provides a set of modules designed to simplify Slurm cluster creation. While it leverages Terraform for underlying infrastructure deployment and management, the Cluster Toolkit offers an abstraction layer that streamlines the operational process, making it a powerful and convenient deployment option.

For detailed information, please refer to the Cluster Toolkit documentation.

NOTE: When deploying with the Cluster toolkit, use the command gcluster deploy <your_config>.yaml. This command manages the underlying infrastructure deployment automatically, replacing the standard Terraform workflow (e.g., terraform init, terraform validat, terraform apply). Please refer to Create the cluster deployment folder for more details.

Terraform

This deployment method directly utilizes Terraform to deploy and manage cluster infrastructure.While potentially more complex than using the Cluster Toolkit, it offers greater flexibility and customization capabilities.

If you are unfamiliar with terraform, we recommend reviewing the official Terraform documentation and the Terraform starter guide to familiarize yourself with its core concepts and operations.

Quickstart Examples

Refer to toolkit_quickstart for an extensible and robust example. This can be configured to handle the creation of all supporting resources (e.g., network, service accounts) or allow you to manage them manually. Slurm can also be configured with partitions and nodesets as desired.

Alternatively,you can find HPC Blueprints within HPC Toolkit examples.