Step 1: Preparing the GCP Deployment

Important: You can currently use the GCP deployment template only with Automation Suite 2024.10. We therefore recommend referring to the Automation Suite 2024.10 documentation.

This page offers insight into the deployment architecture on GCP, the required components, and all known limitations.

Before performing an Automation Suite GCP deployment, you need to ensure you meet the requirements and plan accordingly.

Important:

To prevent data loss, ensure the infrastructure you use does not automatically delete cluster disks on cluster reboot or shutdown. If this capability is enabled, make sure to disable it.

By default, GCP supports only the latest RHEL 8 version. Moreover, it performs automatic updates that may lead to a minor OS version upgrade. As a consequence, you can fall out of the supported Automation Suite OS versions. To mitigate this we recommend using a custom image.

Prerequisites

API requirements

You must have access to a GCP project with the default service account enabled.

This project needs the following APIs enabled, and you must have permissions for all the operations the APIs imply:

• Compute Engine API
• Cloud DNS API
• Cloud SQL Admin API
• Secret Manager API
• Cloud Resource Manager API
• Service Networking API
• Identity and Access Management (IAM) API

To enable an API, take the following steps:

1. Search for the API in the top search bar.

   
   

   

   
   

2. On the Compute Engine API page, select Enable.

   
   

   

   
   
   Note: If you plan to use service account credentials for the deployment, you must enable the Identity and Access Management (IAM) API. The API is used to get the default Compute Engine service account that will be linked to the deployed VMs.

GCP software requirements

You must have GCP SDK and Terraform installed on your machine.

For installation details, see the following:

• Installing Cloud SDK
• Terraform

RSA key pair prerequisites

You need an RSA key pair used for accessing the cluster. You can generate one using the following command:

ssh-keygen -t rsa -b 4096 -C "<ssh_user>"

Quotas

The deployment provisions a configurable number of VMs with configurable types. Additionally, the templates also deploy VMs needed for node registration traffic. Those VMs have a fixed instance type.

Each project has a quota for the number of cores anyone can provision for a specific region.

If the deployment requirements put you over this quota, the deployment would fail. To prevent this, make sure you have enough room in the quota for your Automation Suite deployment.

VM family region availability

Make sure that the VM family region availability meets your requirements.

You can check what VM instances are available in a region at Regions and zones.

(Optional) GPU requirements

Note: Nodes with GPU support are recommended when using AI Center.

Ensure that the GPU you want to use and the region you are deploying satisfy these constraints and the instance type for the GPU nodes supports a GPU. As stated in the GCP documentation, GPUs are currently only supported with general-purpose N1 or accelerator-optimized A2 machine types.

(Optional) Custom image

You can check GCP and RHEL documentation to create a custom image. An alternative is to use Daisy and the workflows provided by GCP.

To avoid automatic updates, you can tie the OS to a specific update using the following command:

subscription-manager release --set=<version>