Choose a control plane deployment
Every cluster has its own control plane. Where that control plane runs is the first decision for a production cluster, and it follows from the infrastructure you already have.
| Deployment | Where the control plane runs | Use it when | License |
|---|---|---|---|
| Kubernetes pod | As pods on an existing Kubernetes cluster | You already run Kubernetes and want to run clusters on it | Open source |
| Standalone binary | As a systemd service on its own Linux nodes | You start from bare metal or VMs with no Kubernetes | None for a single control plane node. High availability needs embedded etcd, which needs the Free tier or higher. |
| Container (vind) | In Docker containers, alongside containerized worker nodes | You need a cluster for local development or CI | None needed |
Kubernetes pod​
The control plane runs as pods on a Kubernetes cluster you already operate, such as EKS, GKE, AKS, OpenShift, or a self-managed cluster. That cluster becomes the control plane cluster, and one control plane cluster can host many clusters.
Worker nodes come from one of two models. Shared nodes are the control plane cluster's own nodes. Private nodes join a single cluster and nothing else. See Decide whether tenants share hardware for that decision.
To deploy, see Deploy on a Kubernetes cluster.
Standalone binary​
vCluster Standalone runs the control plane directly on Linux nodes, with no Kubernetes cluster underneath. It needs its own node for the control plane. You can add control plane nodes later for high availability, and worker nodes join as private nodes.
Standalone supports fewer backing stores than a pod-based control plane. See Choose a backing store.
Standalone also solves the first-cluster problem. To host pod-based clusters on bare metal, deploy Standalone first, then install vCluster Platform on it with Helm. The Standalone cluster becomes the control plane cluster for everything you build on top.
To deploy, see vCluster Standalone.
Container (vind)​
vind runs the control plane and worker nodes as Docker containers, and Docker is its only requirement. Use it for local development, offline CI pipelines, and testing configuration before you deploy a production cluster.
To deploy, see Deploy with Docker.
Decide before you deploy​
Choose the deployment before you create the cluster. Moving to a different deployment means creating a new cluster.
Two other choices are also fixed at creation:
- Worker node model. Shared or private nodes can't change for an existing cluster. See Decide whether tenants share hardware.
- Backing store type. Only a few migrations are supported. See Choose a backing store.
Next steps​
- Choose a backing store for the deployment you picked.
- Size for high availability and scale before production traffic.