Skip to content

KubeVirt Infrastructure Cluster Preparation#

Note

The KubeVirt provider is available starting from k0rdent 1.7.0. See Enable KubeVirt Provider for instructions on how to enable the KubeVirt provider on an existing management cluster after upgrading to 1.7.0.

k0rdent supports using KubeVirt to create and manage clusters backed by KubeVirt virtual machines (VMs). For KubeVirt deployments you must prepare a dedicated infrastructure cluster that will host the KubeVirt VMs (the KubeVirt Infrastructure Cluster). In most cases this cluster is different from the management or regional cluster, which is useful when the management or regional cluster must remain isolated from heavy VM workloads.

Prerequisites#

You can deploy the KubeVirt Infrastructure Cluster by following the instructions in the official KubeVirt documentation. See the Try it out section. Also review the official KubeVirt documentation for more details.

The KubeVirt Infrastructure Cluster must meet the following requirements:

  1. Sufficient Resources The KubeVirt Infrastructure Cluster must have enough resources to host the VMs that will be created. Each worker node should have at least:

    • 8 vCPUs
    • 16 GiB of RAM
    • 120 GiB of disk space

    The actual requirements depend on the size and configuration of the workload clusters.

  2. Nested virtualization (if applicable)

    If the KubeVirt Infrastructure Cluster runs on virtual machines, consider enabling nested virtualization. Follow the instructions described here.

  3. Service LoadBalancer Provisioner

    A Service LoadBalancer provider must be deployed on the KubeVirt Infrastructure Cluster to expose each workload cluster API server via a LoadBalancer Service. For example:

    • Use MetalLB, or
    • When running on a cloud provider, use the Cloud Controller Manager available for that provider.

    The workload cluster API server must be accessible from the management cluster or the regional cluster (when deploying in a region).

  4. Storage Provisioner

    If you plan to use the Containerized Data Importer (CDI) to import VM images for KubeVirt, the KubeVirt Infrastructure Cluster must have a storage provisioner and a StorageClass that supports ReadWriteOnce or better. See Experiment with the Containerized Data Importer for details.

  5. KubeVirt and CDI installed

    Install and configure KubeVirt and Containerized Data Imported (CDI) by following the official documentation:

  6. Create the namespace where the cluster will be deployed

    Due to a known issue in the KubeVirt CAPI provider, the namespace where workload clusters are deployed must be created in advance.

    For example, to create the kcm-system namespace, run:

    kubectl create namespace kcm-system