Skip to content

Overview

This step is a one-time task to setup the shared, multi tenant hardware and software infrastructure. All the end user's compute and services will be deployed on this shared infrastructure.

Overall Steps

Ensure you are logged in as a user with Org Admin credentials. All the steps described below are typically a one-time task.

Info

To provide users with a simple evaluation experience, the steps described below use the Web Console UI. Options for automation are available here.

Step 1: New Project

We will create a new logically isolated environment i.e. a project in our Rafay Org for this guide.

As an administrator, we will want to make sure that all resources we be created in this project. All users we will add will have their access limited to resources in this project.

We will configure and deploy ALL resources in the steps below into this project.

Info

Projects are a unique feature in the Rafay Platform allowing organizations to multi-tenancy i.e. multiple teams, business units and operating environments can be modeled as projects which are completely separate and isolated.

  • Login into your Rafay Org and navigate to Home
  • Click on "New Project", enter a unique name (e.g. gpu-paas) and click on Create

New Project

In a future step in the guide, we will add users to the newly created project using role based access controls in the new project. In addition to role based access, Administrators can also specify policies and enforce resource quotas for the project.

User Access to Project

Step 2: Kubernetes Cluster

Rafay GPU PaaS can be operated on either Rafay provisioned Kubernetes clusters OR clusters imported into Rafay. To make it easy for users to experience the end-to-end workflows of GPU PaaS, we have documented step-by-step instructions on various infrastructure types that may be easily available to users.

Select one of the options below and complete the steps.

# Option
1 Rafay MKS on Linux Server or VM with Real GPUs
2 Rafay MKS on Linux VM on Windows Laptop with Simulated GPUs

Step 3: Deploy Rafay Agent

Infrastructure orchestrated by the Rafay Platform for end users will operate in your private network. The Rafay Platform will never require direct inbound access to any resources in your network. The Rafay Platform uses the Rafay Agent that administrators deploy in their network as a minion to orchestrate the tasks you configure in the Rafay Platform.

In the steps described below, you will deploy an agent onto the newly provisioned Rafay MKS Kubernetes cluster.

  • Navigate to GitOps -> GitOps Agents
  • Click New Agent
  • Enter a name for the agent
  • Select "Kubernetes" for type
  • Select the cluster name
  • Click Create

Create New Agent

In a few minutes, the agent will be deployed onto the cluster. Ensure you wait for the agent to show status as "Healthy" before you move to the next step.

Successful Agent Deployment

Info

Note the name of the agent from this step. We will select this when we configure the templates in a subsequent step.