Overview

Now that you have a management cluster and your SCM connected, you can test out our self-service provisioning using PR Automations and our Terraform management system called Stacks. At a high level, this is going to:

• use a PR Automation (PRA) to instantiate a few CRDs into folders syncable by the root service-of-services in the bootstrap folder
• The InfrastructureStack CRD created by the PRA will create a Terraform stack referencing code in the terraform/modules/clusters/aws folder to provision a new cluster. Your management cluster should already have been configured with sufficient IAM permissions to create this cluster. Any future commits to that folder will also be tracked and generate new Terraform runs to sync in changes to the desired infrastructure.

Enable the cluster-creator PR Automation

There should be a CRD at bootstrap/pr-automation/cluster-creator.yaml, which will create the PRA that drives this tutorial. By default, it references a github SCMConnection CRD, which you'll need to have fully created. Eventually the operator will also create the PR Automation in our API, and it will be visible in the UI as well.

If it's not showing, navigate to the apps service, and you can filter on PrAutomation resources. You should be able to see error messages in the YAML explaining what the operator is tripping on.

Create a Workload Cluster

Now that the PR Automation is configured, we should be able to spawn our cluster seamlessly. The steps are:

• Navigate to https://{your-console-domain}/pr/automations
• Click Create a PR on the cluster-creator Automation Row
  

  

• Fill in the Required Fields
  • Name: The Name of the Cluster
  • Cloud: Cloud Provider to Deploy the Cluster (Dropdown Menu)
  • Fleet: The Fleet to Associate with the Cluster (this is arbitrary but can help you group like clusters)
  • Tier: The Tier to Place the Cluster (dev/staging/prod)

    

• Click Next
• Enter the Name of the Branch to Create the PR

  

• Click Create
  • Optionally view the PR that was created

    

• Merge the PR

Once this PR is merged, and our CD system syncs all the manifests, you should see a clusters service. This will have synced in the InfrastructureStack CRD and caused a Stack to be created at https://{your-console-domain}/stacks.

By default, these stacks require approval for safety (Terraform can do the strangest things sometimes, so you should always validate a Terraform plan before applying). To do that:

• Navigate to https://{your-console-domain}/stacks and click on the run which should be in Pending Approval state
• Click the Approve Button in the top-right
  • You can also see the plan in the run logs and the Plan subtab if you want to ensure the plan looks sane.

Overview of the GitOps Manifests Created

This PRA creates three GitOps manifests:

• A ServiceDeployment at bootstrap/clusters.yaml. This spawns a new service syncing the services/clusters folder. This is meant to reduce bloat in the main bootstrap folder, and is technically optional.
• An InfrastructureStack at services/clusters/{cloud}/stacks/{name}.yaml. This configures the Stacks API to create a new Terraform stack to manage the provisioning of the underlying EKS/AKS/etc cluster. It is tracking the terraform/modules/clusters/{cloud} folder.
• A Cluster CRD at services/clusters/{cloud}/clusters/{name}.yaml. This creates a pointer CRD which is often used in other CRDs like ServiceDeployment to reference clusters.