Single Node OpenShift Lab for Red Hat OpenShift AI on AWS

Description

This project is a basic automation to provision a Single Node OpenShift (SNO) on AWS with a standard installation of Red Hat OpenShift AI and all its dependencies.

By default the SNO is created with the g4dn.metal flavor.

After the basic cluster installation the automation can install and configure the following features:

• HTPasswd identity provider
• Node Feature Discovery Operator
• Nvidia GPU Operator
• Serverless Operator
• Service Mesh Operator
• Red Hat OpenShift AI Operator

Prerequisites

Ansible 2.9+ must be installed on the system. Supported systems are Linux and macOS.

The following additional packages must be installed:

• Python Netaddr to manipulate addressed (python3-netaddr on Fedora/RHEL or netaddr if installing pip3)
• Python OpenShift to access OpenShift API (python3-openshift on Fedora/RHEL or openshift if installing with pip3)
• Python Passlib to generate htpasswd files (python3-passlib on Fedora/RHEL or passlib if installing with pip3)

A future release will automatically handle those dependencies on the different platforms.

The following Ansible Collections are required:

• community.general, used by core modules
• ansible.netcommon, used to apply IP filtering
• kubernetes.core, used to test installation results

To install them:

$ ansible-galaxy collection install -r requirements.yml

Cluster deployment

After receiving the informations run the cluster deploy playbook locally:

$ ansible-playbook main.yaml

Ansible will prompt for sudo password to install the latest oc and openshift-install binaries under the /usr/local/bin path.

Users are expected to provide the following mandatory informations in the user_info.yaml file in order to complete the installation.

• base_domain: the sandbox base domain that will be used to expose APIs and Ingress
• aws_access_key_id: the AWS acces key id available in the received e-mail.
• aws_secret_access_key: the AWS secret access key available in the received e-mail.
• pull_secret: the Red Hat pull secret necessary to pull the cluster images
• ssh_key: the public SSH key that will be injected in the nodes
• install_dir: the installation directory where install files and logs will be created

After the installation login to the cluster using the provided informations in the Ansible output or in the .openshift_install.log file in the installation directory.

Under install_dir/auth the installation deploys the kubeconfig file.

Cluster destroy

The OpenTLC labs use AWS sandboxes for demo and training purposes but it is important to avoid unnecessary compute resource usage, for cost and environmental reasons. After finishing using a cluster destroy it using the follwing command:

$ ansible-playbook destroy.yaml

The install_path variable must be provided (again, in the user_info.yaml file, to confirm the previously used installation directory.

Optional Extra Configs

Users can customize cluster installation configs or day two customizations by editing the cluster_config_vars.yaml. After the base cluster installation is complete users can choose to create custom users with the HTPasswd identity provider or to deploy common extra operators with a simple boolean variable. Available extra configs, with their predefined values, are:

---
# Cluster name
cluster_name: sno

# Sizing of master nodes
master_flavor: g4dn.metal

# AWS install region
aws_region: us-east-2

# The service network used top allocate service vips. To test Submariner
# multicluster connectivity be sure to use diffrenet CIDRs.
service_network: 172.30.0.0/16

# Enable FIPS mode
fips_enabled: false

## Post Install Configs
# HTPasswd Users. Change passwords before installing.
htpasswd_users:
  developer1: RedHat123!
  developer2: RedHat123!

Maintainers

Gianni Salinetti [email protected]