Skip to main content

Kubernetes as a service

Offer your customers managed Kubernetes clusters on your own infrastructure, the way hyperscalers offer services like EKS. Each customer gets a fully isolated cluster with its own control plane and API server. You run the control planes, and by default the customer gets cluster-admin on their own cluster.

This page shows how the pieces fit together and the order to build them in. Each step links to the guide that owns the procedure.

What the service provides​

  • An isolated cluster per customer. Each cluster has its own API server, control plane, and data store, backed by a virtualized control plane that you host.
  • Dedicated worker nodes. With private nodes, a cluster's workloads run only on nodes that belong to that cluster.
  • Self-service provisioning. Customers create clusters from templates you publish, in the platform UI or through the management API.
  • Standard access. Customers download a kubeconfig and use kubectl, Helm, and their usual tooling against their cluster.
  • Capacity on demand. With Auto Nodes, the platform adds and removes worker nodes as each cluster's workloads need them.

Kubernetes is one of the cluster types you can offer from the same infrastructure. The same machines can also back bare metal and VM offerings, and the same templates can add AI platforms to each cluster.

How it works​

LayerWhat it doesWhere it's documented
Control planeRuns each cluster's API server, controllers, and data store, as a pod on your control plane cluster or as a standalone binary on its own machines.Deploy vCluster
Worker nodesRun each cluster's workloads. Private nodes are dedicated to one cluster and run that cluster's own networking.Private nodes
Node supplyAuto Nodes provisions worker nodes through a node provider, such as vMetal's Metal3, NICo, or KubeVirt drivers.Auto Nodes
TemplatesDefine each customer cluster's version, size, node model, policies, and preinstalled apps.Templates
ProjectsDecide who can create clusters, from which templates, and with what quota.Projects

Build the service​

The worker node model you choose in step 2 sets each customer's isolation.

Shared nodes are for trusted tenants

Shared nodes are a supported, common model for trusted tenants. They give each tenant control-plane, API, and namespace isolation, but tenant workloads share the same kernel and physical nodes. They aren't a security boundary for untrusted tenants with Kubernetes access or arbitrary workload execution. Good fits are development, testing, CI/CD, and internal engineering teams.

Use private nodes for external, resale, regulated, or otherwise untrusted tenant offerings, optionally with vNode for runtime isolation.

NetworkPolicy is an added isolation layer worth enabling even for trusted tenants. vCluster can create the policies for you through policies.networkPolicy, and your control plane cluster's CNI enforces them. Confirm your CNI supports enforcement, since some accept NetworkPolicy resources without acting on them. See the security baseline.

If you accept shared nodes for a trusted-tenant use case, follow the shared-node hardening guide as the required next step.

  1. Install vCluster Platform on your control plane cluster. See Install.
  2. Choose the worker node model. Do this before you create any clusters. Private nodes must be enabled when a cluster is created, and an existing cluster can't switch to them. For customers with Kubernetes access, use private nodes. See Decide whether tenants share hardware.
  3. Choose where control planes run. Run them as pods on your control plane cluster, or as standalone binaries when you start from bare machines. See Deploy vCluster and Standalone control planes.
  4. Connect a node supply for private nodes. Create a node provider so Auto Nodes can add worker nodes to each cluster on demand. See Auto Nodes.
  5. Publish your cluster offerings as templates. Set the Kubernetes version, the node model, default network policy, and any apps or Stacks each cluster starts with. See Create templates and Network policy.
  6. Give customers access. Give each customer one or more projects, and allow only the templates their offering includes. See Allowed templates and Quotas.
  7. Let customers create clusters. Customers create a cluster from a template and download its kubeconfig. To test the service, create one cluster from a customer project and run kubectl get nodes against it. See Create a cluster from a template and Retrieve a kubeconfig.

To put your own product in front of the platform, drive cluster creation from your product's API instead of the platform UI. See Build a product control plane on the Platform API.

Operate the service​

Know where the service ends​

You own the control planes, the control plane cluster that hosts them, the node supply, and upgrades. The customer owns what runs inside their cluster, including their workloads, the namespaces and RBAC they create, and any operators they install. Decide which add-ons you manage for customers, and ship those through templates so every cluster starts with them.