Feature Overview
This feature covers the creation of a small-scale, validated Red Hat OpenStack Platform (RHOSP) 18 cluster. This cluster will be specifically designed to simulate a production environment for the purpose of testing and validating hardware accelerators (like GPUs) for AI/ML workloads. The initial deployment will be in the BRQ2 location and will focus on validating the end-to-end customer experience.

Goals

• Deploy a functional, stable, small-scale RHOSP 18 cluster in the BRQ2 lab.

• Establish a repeatable, automated process for deploying and managing the cluster to ensure consistency.

• Create environment suitable for testing and benchmarking various AI accelerator hardware.

• Enable engineering teams to effectively simulate and validate customer AI workload scenarios.

Requirements
Hardware Provisioning:

• Identify and secure sufficient physical servers in the BRQ2 lab to act as control, compute, and storage nodes.

• Ensure the selected hardware meets the minimum requirements for RHOSP 18 and the specific AI accelerators to be tested.

Network Configuration:

• Design and implement the necessary network infrastructure, including VLANs, subnets, and routing.

• The network must support all required RHOSP traffic types (Control, Internal API, Storage, Tenant, External).

Deployment Automation:

• Develop robust automation (e.g., using Ansible) for the bare-metal provisioning and deployment of the complete RHOSP 18 cluster.

• The automation must be idempotent, configurable, and version-controlled.

Acceptance Criteria
[ ] A RHOSP 18 cluster is successfully deployed and fully operational in the BRQ2 location.

[ ] All core OpenStack services (Keystone, Nova, Neutron, Cinder, Glance) are healthy, accessible via API/CLI, and pass health checks.

[ ] The entire deployment process is automated. The cluster can be torn down and redeployed from scratch using the automation with minimal manual intervention.
[ ] At least one model of AI accelerator can be successfully provisioned to a Nova instance (e.g., via PCI-passthrough) and is usable from within the guest OS.