OCI HPC/GPU Cluster Stack for Slurm

The OCI HPC/GPU Cluster Stack deploys and operates CPU and GPU high-performance computing clusters on Oracle Cloud Infrastructure (OCI). It combines Terraform, OCI Resource Manager, Ansible, Slurm, and OCI-native services to:

• Provision repeatable HPC infrastructure in a new or existing VCN.
• Schedule CPU and GPU workloads with Slurm, including optional high availability and federation.
• Scale clusters through cluster networks, compute clusters, instance pools, and standalone nodes.
• Accelerate workloads with RDMA, NVIDIA GPUs, AMD GPUs, and GPU memory-fabric configurations.
• Integrate OCI File Storage, Lustre, local NVMe, and Object Storage.
• Operate clusters with lifecycle automation, monitoring, health checks, alerting, and the mgmt management CLI.

Compatibility

Component	Supported baseline
Terraform CLI	Version 1.2 or later
OCI Terraform provider	Version 7.19.0 or later
Operating systems	See prerequisites
Compute images	See image guidance
Compute shapes	Supports OCI compute shapes compatible with the selected image and OCI region.
Deployment methods	OCI Resource Manager or Terraform CLI

Compatibility also depends on the selected region, availability domain, service limits, image, compute shape, network, and storage configuration.

Installation

OCI Resource Manager is the recommended deployment method. Use the deployment button above or upload a release archive to Resource Manager. Terraform CLI deployments are also supported when OCI authentication and all required input variables are provided externally.

Deployment steps:

1. Review the prerequisites.
2. Configure the stack using the setup guide.
3. Deploy the stack through Resource Manager or Terraform CLI.
4. Use the management CLI to validate and operate the resulting cluster.

Warning

This stack creates chargeable OCI resources. Review the Terraform plan and estimate the expected cost with the Oracle Cloud Cost Estimator before applying it.

Documentation

• Getting started
  • Documentation overview
  • Prerequisites and supported images
  • Stack configuration and deployment
  • Stack user guide
• Architecture and configuration
  • Architecture overview
  • Sequence diagram: Terraform provisioning, node bootstrap, event, and management flows
  • Cluster Lifecycle
  • Monitoring and alerting
  • GB200/GB300 memory-fabric deployments
  • DGXC benchmarking
• Operations and runbooks
  • Cluster management CLI
  • Instance maintenance events and policy
  • Create a Slurm GPU partition
  • Slurm troubleshooting
  • Run health checks and NCCL validation
  • Recover local NVMe storage
  • Centralized logging and troubleshooting
• Reference
  • Slurm monitoring metrics decision record
  • Package versioning decision record

Examples

See the samples/ directory for example configuration and workload artifacts. Operational procedures, validation, and recovery steps are in the documentation overview.

Help

• Open a GitHub issue for bug reports, questions, or enhancement requests.
• Do not include credentials, private keys, authentication tokens, complete OCIDs, private addresses, or unsanitized logs in public issues.

Contributing

Contributions from the community are welcome. Before submitting a pull request, review the contribution guide.

Security

Report security vulnerabilities according to the security guide. Do not report vulnerabilities through a public GitHub issue.

License

Copyright (c) 2018-2026 Oracle and/or its affiliates.

Released under the Universal Permissive License v1.0 as shown at https://oss.oracle.com/licenses/upl/.