> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rafftechnologies.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Upgrade the Kubernetes version

> Move a cluster to a newer Kubernetes version with a rolling in-place upgrade

<sub>Updated August 24, 2026</sub>

Clusters upgrade in place: control-plane nodes one at a time, then workers drained one at a time. On an HA cluster the Kubernetes API stays available throughout; on a single-master cluster the API is briefly interrupted while the master restarts. Workloads with more than one replica keep serving either way. Upgrades are free and cannot be rolled back.

## Steps

<Steps>
  <Step title="Open Settings">
    On the cluster page, open **Settings** and find **Kubernetes version**. The card shows your current version and the available targets — minor versions are offered sequentially (1.33 → 1.34, no skipping).

    <Frame>
      <img src="https://mintcdn.com/rafftechnologiesllc/kqSIuas76DRL-V4V/images/products/compute/kubernetes/version-card.png?fit=max&auto=format&n=kqSIuas76DRL-V4V&q=85&s=845cafbbeefbda808abc4a57aa639b03" alt="Kubernetes version card in cluster Settings — current version, upgrade mode, and maintenance window" width="1612" height="511" data-path="images/products/compute/kubernetes/version-card.png" />
    </Frame>
  </Step>

  <Step title="Choose a version and upgrade">
    Pick the target version and click **Upgrade**. Confirm the dialog — on a single-master cluster it warns about the brief API interruption.
  </Step>

  <Step title="Watch the progress">
    The header shows an **Upgrading** badge and the **Activity** feed reports each node as it moves to the new version ("2 of 5 nodes upgraded"). A 5-node cluster typically takes 5–15 minutes.
  </Step>
</Steps>

## Verify

When the badge clears, the cluster header shows the new version, and `kubectl get nodes` lists every node on it:

```bash theme={null}
kubectl get nodes
```

## Automatic upgrades

In the same Settings card you can switch the upgrade mode:

* **Manual** (default) — you start every upgrade yourself.
* **Auto-patch** — patch releases (e.g. 1.34.6 → 1.34.9) apply automatically in your weekly maintenance window, with a notice email \~72 hours before.
* **Auto-minor** — new minor versions too, after a stability period.

The maintenance window is a 4-hour weekly slot (day + start hour, UTC). It is not downtime — upgrades are rolling and workloads keep running.

<Note>
  When a version reaches end of life, remaining clusters are upgraded automatically in their maintenance window regardless of mode. You are emailed 60 and 14 days before. See [Supported versions](/products/build/kubernetes/details/supported-versions).
</Note>