Red Hat OpenShift AI Administration

Welcome to this quick course on the Red Hat OpenShift AI Administration. This course is the second in a series of five courses about Red Hat OpenShift AI:

  1. Introduction to Red Hat OpenShift AI

  2. Red Hat OpenShift AI Administration (This course)

  3. Creating Machine Learning Models with Red Hat OpenShift AI

  4. Deploying Machine Learning Models with Red Hat OpenShift AI

  5. Automation using Data Science Pipelines

After you have completed all the courses in the learning path, you can attempt the "Workshop final hands-on lab", that tests your understanding of the concepts taught in all the courses.

Authors

The PTL team acknowledges the valuable contributions of the following Red Hat associates:

  • Jiri Tesar

  • Jaime Ramirez Castillo

  • Max Murakami

  • Trevor Royer

  • John Hurlocker

  • Ravi Srinivasan

Classroom Environment for Red Hat Associates

Red Hat associates should use the Red Hat Demo Platform to provision a classroom.

Use the Red Hat OpenShift Container Platform Cluster catalog item from Red Hat Demo Platform (RHDP).

When ordering this demo item:

  • Select Practice/Enablement for the Activity field, and Learning about the product for the Purpose.

  • If you have a Salesforce ID for a customer sales opportunity, then copy it to the Salesforce ID field. Otherwise, add some random text like "Learning RHOAI".

  • Select 5 users in the User Count field. Having at least two users is required for the user management exercise.

  • In the OpenShift worker memory size field, select 64Gi. This size might be required if you run multiple workbenches and pipelines simultaneously.

  • In the OpenShift version field, select the latest stable version available.

  • Review the auto-stop and auto-destroy timers. You can change this later after classroom provisioning by opening a ticket with the RHDP team.

RHOAI workbenches and pipelines use large container images. Dependencies such as Tensorflow, PyTorch, or CUDA result in image sizes of several gigabytes.

The data ingestion and training stages can demand large amounts of memory and CPU/GPU resources. Make sure that your worker nodes have enough memory to accommodate pods that consume dozens of GB.

This classroom does NOT have RHOAI pre-installed. You install and configure a basic RHOAI instance, and then continue using this instance with other courses in the learning path.

Alternative classroom environment on AWS for Red Hat Associates

You can also use an AWS cluster by ordering the AWS with OpenShift Open Environment.

This classroom does NOT have RHOAI pre-installed. You install and configure a basic RHOAI instance, and then continue using this instance with other courses in the learning path.

If you use this item, you must:

  • Select Practice/Enablement for the Activity field, and Learning about the product for the Purpose.

  • If you have a Salesforce ID for a customer sales opportunity, then copy it to the Salesforce ID field. Otherwise, add some random text like "Learning RHOAI".

  • Make sure that you select 3 as the control plane count and the latest OpenShift version.

  • Select your nearest AWS region and choose m6a.4xlarge as the instance type (Ensure that whatever instance type you choose has at least 16 vCPUs and 64GB RAM).

  • Review the auto-stop and auto-destroy timers. You can change this later after classroom provisioning by opening a ticket with the RHDP team.

  • Configure users and authentication. The AWS item is only pre-configured with the kubeadmin user, which does not have the cluster-admin role assigned. The exercises in this course, however, assume the existence of a user named admin with the cluster-admin role and five non-admin users (user1…​user5). Download the following two scripts to your local machine, and execute the following commands.

The password for both the admin and user1…​user5 users is openshift23.

  • You may have to wait for 5 minutes while the changes take effect. You will see a log in prompt with the htpasswd_provider displayed.

    htpasswd provider
    Figure 1. htpasswd_provider prompt

  • Click the htpasswd_provider button, and log in as the admin user with password openshift23 to access the OpenShift web console as a cluster administrator.

Classroom Environment for Red Hat Partners

Red Hat partners should provision their own Red Hat OpenShift cluster using the Red Hat Hybrid Cloud Console at https://console.redhat.com/openshift/overview.

  • Ensure that you provision an OpenShift 4.13 or later cluster with a minimum of 3 control plane nodes (each with 8vCPU and 16GB RAM), and 3 worker nodes (each with 16 vCPU, and 64GB of RAM).

  • Configure users and authentication. A default OpenShift cluster installation has a user called kubeadmin. The exercises in this course, however, assume the existence of a user named admin with the cluster-admin role and five non-admin users (user1…​user5). Download the following two scripts to your local machine, and execute the following commands after logging in to your cluster as the kubeadmin user using the oc CLI tool.

The password for both the admin and user1…​user5 users is openshift23.

  • You may have to wait for 5 minutes while the changes take effect. You will see a log in prompt with the htpasswd_provider displayed.

    htpasswd provider
    Figure 2. htpasswd_provider prompt

  • Click the htpasswd_provider button, and log in as the admin user with password openshift23 to access the OpenShift web console as a cluster administrator.

Prerequisites

  • Basic knowledge of OpenShift (or Kubernetes) administration

  • Ability to build and deploy container images

  • Knowledge of OpenShift User and Role administration

  • Basic knowledge of AWS EC2 and S3 services

Objectives

The overall objectives of this course include:

  • Install RedHat OpenShift AI using the web console and CLI

  • Upgrade RedHat OpenShift AI components

  • Manage RedHat OpenShift AI users and controlling access

  • Enable GPU support in RedHat OpenShift AI

  • Stop idle notebooks

  • Create and configure a custom notebook image