rhoai-demo-setup

Summary

This repository provides automated setup and configuration tools for deploying Red Hat OpenShift AI (RHOAI) on OpenShift clusters. It's designed to streamline the process of setting up a complete AI/ML platform with GPU support and related infrastructure.

Key Purposes:

1. Automated RHOAI Deployment: Uses Ansible playbooks to automate the installation and configuration of Red Hat OpenShift AI and all its dependencies on OpenShift clusters.

2. GPU-Enabled AI Infrastructure: Sets up GPU nodes with NVIDIA operators, drivers, and monitoring dashboards to support AI/ML workloads that require GPU acceleration.

3. Complete Platform Setup: Installs and configures supporting components including:

   • Serverless and Service Mesh operators
   • Authentication and authorization (Authorino)
   • Object storage (MinIO)
   • Monitoring and observability tools
   • Web terminals and developer tooling

4. Demo Capabilities: Provides ready-to-use demos showcasing RHOAI capabilities, particularly model serving with vLLM.

Main Components:

• components: Kustomize manifests for 16 different cluster components (prerequisites, GPU operators, RHOAI, monitoring, etc.)
• playbooks: Ansible automation scripts for full cluster setup, individual component installation, and demos
• demos: Sample applications and inference examples
• docs: Documentation and setup guides

Primary Use Cases:

• Full Cluster Setup: One-command deployment of a complete RHOAI environment
• Individual Component Setup: Granular installation of specific components (GPU setup, MinIO, etc.)
• AI/ML Demonstrations: Pre-configured demos for model serving and inference

This repository essentially serves as a "one-stop-shop" for getting a production-ready AI/ML platform running on OpenShift with minimal manual configuration required.

Pre-Requisites

Before using this repository, ensure you have the following:

1. OpenShift Cluster

• You must have access to a running OpenShift cluster (Instructions on creating a cluster)
• You need the kubeadmin credentials for cluster administration.

2. Ansible and ansible-playbook

• For macOS:

  brew install ansible

• For Linux:

  • Using pip (if Python is installed):

    pip install --user ansible

  • For Debian/Ubuntu:

    sudo apt update
    sudo apt install ansible

  • For RHEL/CentOS/Fedora:

    # For Fedora
    sudo dnf install ansible
    # For RHEL/CentOS
    sudo yum install ansible

• For Windows:

  • Using pip (if Python is installed):

    pip install --user ansible

  • Using WSL (Windows Subsystem for Linux):

    # Install Ansible inside WSL (Ubuntu/Debian)
    sudo apt update
    sudo apt install ansible

• Verify installation:

  ansible --version
  ansible-playbook --version

3. OpenShift CLI (oc)

• Install the OpenShift CLI

  • Official documentation.
  • Download for Windows Link
  • Download for Mac LINK
  • Download for Linux LINK

• Verify installation:

  oc version

---

Automation

Below are the main Ansible files under the playbooks/ directory (top-level only) that get run as part of the automation.

Full Cluster Setup

Run with:

ansible-playbook playbooks/cluster-setup.ansible.yml

• Below components get installed:

  • Install cluster pre-reqs
  • Add user with cluster-admin role
  • Provision and configure GPU node
  • Install NFD and NIVIDIA GPU operator
  • Install Serverless and Servicemesh operator
  • Install RHOAI and depdendent components

Important

When running this playbook, you don't need to run automation for individual components as they get run as part of full cluster setup.

Individual Component Setup

• Provision and configure GPU node

  Run with:

  ansible-playbook playbooks/gpu-setup.ansible.yml

  • Below components get installed:
    • Add GPU node
    • Install NFD and NVIDIA GPU operator
    • Install NVIDIA DCGM dashboard
    • Configure timeslicing in GPU
    • Taint GPU nodes

• Provision MINIO object storage

  Run with:

  ansible-playbook playbooks/minio-setup.ansible.yml

Demos

• Demo model serving on vLLM

  Run with:

  ansible-playbook playbooks/demo-vllm.ansible.yml

  • Below components get installed:
    • Install MINIO storage
    • Create DataScience project
    • Create data-connections
    • Create ServingRuntime
    • Create InferenceService

Instructions

1. Clone the Repository

git clone https://github.com/redhat-ai-americas/rhoai-demo-setup.git
cd rhoai-demo-setup

2. Run the Ansible Playbooks

You can run any playbook using the following command:

ansible-playbook <path-to-playbook>

For example, to run the full cluster setup:

ansible-playbook playbooks/cluster-setup.ansible.yml

Or to run a specific setup or demo playbook:

ansible-playbook playbooks/gpu-setup.ansible.yml
ansible-playbook playbooks/minio-setup.ansible.yml
ansible-playbook playbooks/demo-vllm.ansible.yml

Note

To save the output to a log file, use:

ansible-playbook playbooks/cluster-setup.ansible.yml | tee cluster-setup.log

You can refer to github repo demo-ai-gitops-catalog for for more in-depth RHOAI installation and customization.