Skip to main content
Version: v0.38 Stable

Plan your cluster

This section covers the decisions to make before a cluster goes to production. It doesn't install anything. Each page explains one choice, the trade-offs behind it, and the Deploy or Configure page that carries it out. Read it before Deploy, whether you're building one production cluster or a platform that hands clusters to others.

Planning comes first because some choices are fixed when a cluster is created. A cluster that starts on shared nodes can't move to private nodes later, and changing the backing store type in place isn't supported. Getting those right up front saves a rebuild. Other choices, such as sizing, can be adjusted after the cluster is running.

Make the decisions in order​

Each decision narrows the next, so work through them in this order.

  1. Choose a control plane deployment. Decide whether the control plane runs as a pod on a Kubernetes cluster, as a Standalone binary on its own nodes, or in Docker. This limits the choices that follow. For example, vCluster Standalone uses only private nodes.
  2. Decide whether tenants share hardware. Choose between shared and private worker nodes. This sets the tenant isolation boundary, and it can't change after creation.
  3. Choose a backing store. Pick where the cluster stores its state. The options depend on the deployment you chose, and only some migrations are possible after you deploy.
  4. Size for high availability and scale. Set the replica count, resource profile, DNS capacity, and control plane cluster capacity for the load you expect.
  5. Production readiness checklist. Confirm the cluster is ready before it takes production traffic, from hardening to backups, upgrades, and monitoring.

Plan around failure​

Two pages explain behavior to design around rather than a choice to make. What happens during control plane outages? describes how a cluster behaves when its own control plane or the control plane cluster is unavailable. Why does CoreDNS become a DNS bottleneck? explains when a single CoreDNS replica runs out of headroom, which feeds into the sizing decision above.

When the decisions are made, deploy the cluster and tune the settings they lead to in Configure. If you're new to vCluster, start with What is vCluster? for how its architecture works.