Deploy

Use this deployment flow to provision and validate a production-ready HPC and AI platform on OCI with OOD, Slurm, identity, and storage services.

Follow these steps:

  1. Create the OCI Networking networking environment.
  2. Deploy shared storage services.
  3. Configure FreeIPA identity services.
  4. Deploy Slurm controllers and login nodes.
  5. Configure autoscaling CPU and GPU partitions.
  6. Build or import CPU and GPU golden images by using a controlled image pipeline.
  7. Deploy Open OnDemand.
  8. Integrate Microsoft Entra ID OIDC authentication.
  9. Install and verify Batch Connect applications.
  10. Configure monitoring and alerting.
  11. Verify storage, scheduler, autoscale, and interactive applications from browser and command-line workflows.
  12. Capture post-upgrade and rollback runbooks.
  13. Promote only verified configurations and scripts back into source control.

You have deployed and validated the platform across infrastructure and user workflow paths, with runbooks and controlled configuration baselines for operations.

Production Readiness Checklist

Use this checklist to confirm user-facing and infrastructure readiness before production rollout of an OOD-based HPC and AI platform on OCI.

Before production rollout, verify the architecture from both the infrastructure layer and the user workflow layer.

Recommended readiness checks include OOD web backend redundancy, OIDC routing behavior, Slurm partition coverage, Remote Desktop proxy routing, autoscale bootstrap validation, GPU runtime validation, shared runtime consistency, and post-upgrade compatibility checks.

Priority Item Reason
Critical Authenticated browser acceptance Confirm real browser flows after login for Desktop, Code Server, JupyterLab, Shell, Files, and Job Composer.
Critical GPU Lustre route decision Define which GPU image lane supports native Lustre and expose the choice in partition names and user guidance.
High GPU golden image standardization Reduce cold-start time, user and group drift, scheduler runtime drift, and repeated node-side repair.
High Shared interactive runtime governance Version and validate shared runtimes under common paths such as /apps across partitions.
High Project and lab group governance Enforce multi-tenant access control, quota ownership, and chargeback or showback operations.
High Automated OOD post-upgrade patching Prevent repeat regressions after future OOD upgrades.