By default, the installation program installs control plane and compute machines into a single Nutanix Prism Element (cluster). To improve the fault tolerance of your OpenShift Container Platform cluster, you can specify that these machines be distributed across multiple Nutanix clusters by configuring failure domains.

A failure domain represents an additional Prism Element instance that is available to OpenShift Container Platform machine pools during and after installation.

Failure domain requirements

When planning to use failure domains, consider the following requirements:

  • All Nutanix Prism Element instances must be managed by the same instance of Prism Central. A deployment that is comprised of multiple Prism Central instances is not supported.

  • The machines that make up the Prism Element clusters must reside on the same Ethernet network for failure domains to be able to communicate with each other.

  • A subnet is required in each Prism Element that will be used as a failure domain in the OpenShift Container Platform cluster. When defining these subnets, they must share the same IP address prefix (CIDR) and should contain the virtual IP addresses that the OpenShift Container Platform cluster uses.

Installation method and failure domain configuration

The OpenShift Container Platform installation method determines how and when you configure failure domains:

  • If you deploy using installer-provisioned infrastructure, you can configure failure domains in the installation configuration file before deploying the cluster. For more information, see Configuring failure domains.

    You can also configure failure domains after the cluster is deployed. For more information about configuring failure domains post-installation, see Adding failure domains to an existing Nutanix cluster.

  • If you deploy using infrastructure that you manage (user-provisioned infrastructure) no additional configuration is required. After the cluster is deployed, you can manually distribute control plane and compute machines across failure domains.