×

In OKD version 4.17, you can install a three-node cluster on Microsoft Azure. A three-node cluster consists of three control plane machines, which also act as compute machines. This type of cluster provides a smaller, more resource efficient cluster, for cluster administrators and developers to use for testing, development, and production.

You can install a three-node cluster using either installer-provisioned or user-provisioned infrastructure.

Deploying a three-node cluster using an Azure Marketplace image is not supported.

Configuring a three-node cluster

You configure a three-node cluster by setting the number of worker nodes to 0 in the install-config.yaml file before deploying the cluster. Setting the number of worker nodes to 0 ensures that the control plane machines are schedulable. This allows application workloads to be scheduled to run from the control plane nodes.

Because application workloads run from control plane nodes, additional subscriptions are required, as the control plane nodes are considered to be compute nodes.

Prerequisites
  • You have an existing install-config.yaml file.

Procedure
  1. Set the number of compute replicas to 0 in your install-config.yaml file, as shown in the following compute stanza:

    Example install-config.yaml file for a three-node cluster
    apiVersion: v1
    baseDomain: example.com
    compute:
    - name: worker
      platform: {}
      replicas: 0
    # ...
  2. If you are deploying a cluster with user-provisioned infrastructure:

    • After you create the Kubernetes manifest files, make sure that the spec.mastersSchedulable parameter is set to true in cluster-scheduler-02-config.yml file. You can locate this file in <installation_directory>/manifests. For more information, see "Creating the Kubernetes manifest and Ignition config files" in "Installing a cluster on Azure using ARM templates".

    • Do not create additional worker nodes.

Example cluster-scheduler-02-config.yml file for a three-node cluster
apiVersion: config.openshift.io/v1
kind: Scheduler
metadata:
  creationTimestamp: null
  name: cluster
spec:
  mastersSchedulable: true
  policy:
    name: ""
status: {}