How paasbox works

paasbox is a thin, opinionated software layer over Gardener, the open-source Kubernetes-cluster manager from SAP. You connect your own Hetzner account and get a managed Kubernetes platform: an automated control plane that provisions and operates a production cluster — while the worker nodes run in your Hetzner project, at your price.

The shape: Garden, Seed, Shoot

Gardener is “Kubernetes that runs Kubernetes”, in three layers:

• A Garden cluster is the brain of the whole system — it holds the configuration and orchestrates everything else.
• Seed clusters host the control planes of your workload clusters. This is the clever part: your control plane runs as ordinary pods inside a seed, virtualized away from your own machines. It’s highly available, resource-efficient, and lets a cluster hibernate cleanly.
• Shoot clusters carry your actual workloads. A shoot is, in practice, a YAML document that Gardener reconciles into a real cluster.

This architecture is battle-tested at a scale most platforms never reach — it powers SAP’s own Kubernetes service, STACKIT’s cloud, and T-Systems’ sovereign cloud, with people who have operated fleets of thousands of clusters for years. With paasbox, you get the Garden and the seeds as an automated platform — delivered entirely as software.

Who runs what

The split is the whole point:

• We operate the Garden and the seeds — so your cluster’s control plane (API server, etcd, scheduler, controllers) runs as managed pods on our seed clusters. You never host, size, patch, or pay per-node for a control plane.
• You bring a Hetzner Cloud project and an API token. Your worker nodes, load balancers, and volumes live in your own Hetzner project, billed to you directly by Hetzner — ideally at your grandfathered rate.

See What you bring & what we run for the full table.

The path to a running cluster

1. Sign up & add a payment method

   Create an account, join (or start) a team, and put a card on file. The card gates cluster creation — it’s what lets us run your control plane without it ever running unpaid. There’s no charge until you start a cluster.

2. Connect Hetzner

   Paste a Hetzner Cloud API token (read/write). paasbox validates it, stores it encrypted, and wires it into the platform. (Why read/write?)

3. Adopt the servers you want to keep

   paasbox lists your existing Hetzner servers; you pick which to bring under management. Adoption rebuilds each server to put it under Kubernetes — so migrate your workloads off first. From then on, those servers are rebuilt in place and never deleted, which is how the grandfathered price survives every upgrade.

4. Configure the cluster

   Choose a name, a Kubernetes version, a region, and your node pools — a pooled pool over your adopted (grandfathered) servers, plus an optional managed pool of on-demand nodes for burst. paasbox defaults to pooled-first so you get the savings without tuning anything.

5. Start it

   paasbox opens the per-cluster subscription (prorated) and creates the shoot, then shows live reconciliation until the cluster is Ready.

6. Download your kubeconfig

   Pull a short-lived admin kubeconfig whenever you need one. It’s a real cluster-admin credential — not a locked-down console.

7. Run & operate

   Deploy and manage your workloads directly on the cluster with kubectl, Helm, or Flux. In paasbox you watch metrics, availability, capacity, and cost/savings, and manage lifecycle: scale, upgrade, hibernate, add or remove pooled servers, or delete the cluster (which stops the charge, prorated).

The two kinds of node pool

• Pooled (grandfathered) — worker nodes adopted from your pre-increase servers. Ranked first by the autoscaler, rebuilt in place on every operation, never deleted. This is your cheap baseline.
• Managed (on-demand) — ordinary Gardener-provisioned nodes at the current price, used for burst above the baseline and able to scale to zero. The autoscaler only spills here when the pool is full.

You mix both in one cluster and pay the same flat fee either way.