---
title: "Versions and upgrades"
description: "Supported Kubernetes versions, automatic patch updates, minor upgrades and end of support."
---

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

# Versions and upgrades

## Versions

Clusters run Kubernetes 1.36 or 1.37. Each minor version is supported for 14 months from its upstream
release; there's no extended support. New minor versions are added after they pass the CNCF conformance
tests.

`GET /public/v1/versions` lists the versions offered, each with its current patch version and end of
support date (`ezgh kubernetes versions list`). A new cluster gets the newest unless you pick one.

## Patch updates

EZGH applies patch releases to the control plane for you, during the cluster's weekly maintenance
window: 4 hours from `startHour` (UTC) on `day`. EZGH picks a window when you don't.

```sh
ezgh kubernetes clusters update kcl_3k9d0a1m2x7q --maintenance-window sunday:9 --wait
```

## Minor upgrades

You upgrade a cluster yourself, one minor version at a time and never down:

```sh
ezgh kubernetes clusters update kcl_3k9d0a1m2x7q --version 1.37 --wait
```

The API keeps serving while the control plane is upgraded. Add-ons that aren't pinned move to the new
version's defaults.

Before anything changes, the upgrade is checked, and refused with `409 precheck_failed` and a list of
`reasons`, each a `code` and a `message`, when:

| Code | When | The message names |
| --- | --- | --- |
| `node_group_skew` | A node group would be more than 3 minor versions behind the cluster | The node group |
| `addon_unsupported` | A pinned add-on's version doesn't run on the new version, or an add-on has no version for it | The add-on |
| `removed_api_in_use` | The cluster still serves requests through an API the new version removes | Each API's group, version and resource |

Then upgrade each node group to the cluster's version. Its nodes are replaced a few at a time, each
drained first ([Node groups](/kubernetes/node-groups#upgrades-and-draining)).

## End of support

60 days before a version's end of support, the cluster records an `end_of_support` [event](/kubernetes-api/ListClusterEvents/), and the
console shows the date on the cluster. After end of support, the cluster is upgraded to the next minor
version in its first maintenance window. If the pre-checks refuse that upgrade, an event says why, and
it's tried again in each maintenance window after; it's never forced.

Source: https://docs.ezghcloud.com/kubernetes/upgrades/index.mdx
