Cluster Stacks

Introduction

A Cluster Stack is a versioned bundle of tested components that defines how a Kubernetes cluster is built and managed by Syself Autopilot. It includes the Kubernetes version, CNI, CSI, OS image, and all the components needed to provision a production-grade cluster. You can think of it as a Dockerfile for your cluster.

Each Cluster Stack provides a deterministic, declarative, and reproducible cluster configuration — ensuring your cluster behaves as expected, every time. This also allow us to heavily test and validate each Cluster Stack version before releasing it.

Advantages of Cluster Stacks

With a Cluster Stack, you're not building a cluster from dozens of moving parts. You're selecting a tested, versioned package that Syself has validated end-to-end — including upgrade paths.

This approach gives you:

  • Predictable upgrades: Each new Cluster Stack version is tested extensively before release, and all upgrade paths are tried and documented.
  • Faster provisioning: Cluster creation is faster because all components are bundled and pre-configured with sane defaults.
  • Simplified operations: You don’t have to assemble or maintain cluster components. We provide a consistent baseline, and the system ensures it stays healthy.
  • Full infrastructure control: Although the configuration is packaged, everything still runs on your infrastructure — nothing is hidden or locked away from you.

Simple, Safe Upgrades

Kubernetes upgrades aren’t as simple as bumping a version number. Even minor version upgrades often require changes to multiple components — and the order matters. For instance, your CNI plugin might need to be upgraded before the control plane, while CSI drivers and OS images may have their own sequencing constraints. Missing a step or applying them in the wrong order can lead to downtime or broken clusters.

By simply selecting a new Cluster Stack version, you're triggering a carefully orchestrated upgrade process. The system knows exactly which components need to be updated, in what order, and under which conditions. Every step is automated, validated, and designed to maintain stability.

You never have to manually coordinate updates across networking, storage, control plane, or OS layers. Just pick the version — we handle the rest.

How it Works

Each Cluster Stack has a name like hetzner-apalla-1-32-v2 . This refers to:

  • The provider: hetzner
  • The Kubernetes version: 1.32
  • The Cluster Stack version: v2

When you create a workload cluster, you select the latest supported Cluster Stack for that environment. Everything from the Kubernetes control plane to the OS image and storage driver is provisioned based on this versioned stack.

Although you can’t customize the internals of a stack, they’re designed with flexibility in mind. For example, if we detect a GPU during provisioning, drivers and runtime components are automatically installed.

Upgrading

Upgrading your cluster is as simple as selecting a new Cluster Stack version. Behind the scenes, Syself Autopilot handles the orchestration of component upgrades safely and incrementally, following the paths we've already validated.

You don’t need to worry about individual components being out of sync — upgrades are atomic at the stack level.

Previous
Self-Healing
Next
Management and Workload Clusters