Glossary

Every product noun PaaSbox uses, defined in one place. Terms link to the page that covers them in depth.

Cluster — a complete managed Kubernetes cluster: a control plane we operate plus the worker nodes in your Hetzner project. The unit you create, hibernate, delete — and pay for.

Control plane — the brain of a cluster: API server, etcd, scheduler, controllers. On PaaSbox it runs as managed components on platform infrastructure — never in your Hetzner project, never on your Hetzner bill. Comes in two tiers: Basic (single replica) and HA (3× etcd and API server).

Node pool — a group of worker nodes with one shape and one source. A cluster has one or more; the source — Managed or Pooled — decides how the pool behaves.

Managed pool — worker nodes the platform orders on demand in your Hetzner project, at current prices. Cattle: resizable, replaceable, able to scale to zero.

Pooled pool — worker nodes drawn from your fleet of adopted servers. Pets: rebuilt in place, capped at the fleet, never deleted — the pool that preserves grandfathered prices.

Adopt / adoption — bringing a Hetzner server you already own under PaaSbox management: the server is wiped and reinstalled in place (typed WIPE confirmation), joins your fleet, and is never deleted from then on. See adopt your servers.

Fleet — your team’s adopted servers, grouped by server type and location, whether currently claimed by a cluster or not. Pooled pools draw from it; hibernating releases servers back to it.

Grandfathered — a Hetzner server ordered before 15 June 2026, 08:00 CEST that keeps its pre-increase price for as long as that exact instance is never deleted or rescaled. The reason PaaSbox exists — the full story.

Hibernate / wake — putting a cluster to sleep and bringing it back: on-demand nodes removed, pooled servers released to the fleet, control plane scaled to zero, all state retained. On the hourly plan a hibernated cluster meters €0. See hibernation.

Connection — a stored credential to your infrastructure account: today, a Hetzner Cloud API token, scoped to one project, held encrypted and never displayed again.

Region — where a cluster lives: control plane and worker nodes in the same Hetzner location. nbg1 is live; others open at demand — see regions.

kubeconfig — the credential kubectl uses. PaaSbox issues short-lived admin kubeconfigs (up to 8 hours, minted on demand, issuance audited) for your cluster — access your cluster.

Team — the ownership unit for everything: clusters, connections, fleet, billing. Two roles, Administrator and Member; no per-seat charges. See teams & API keys.

API key — a durable secret (pbx_live_…) for scripts and CI, created in the console, shown once, revocable. Used to fetch disposable kubeconfigs per run.

PAYG / awake hours — the hourly billing plan: the meter counts hours your cluster is awake, partial hours round up, each active period bills at least 4 hours, hibernated time meters €0. See pricing & billing.

Spend cap — the hard monthly ceiling on your team’s PaaSbox spend, from €200 on a new account upward with billing history. At the cap, clusters are hibernated — never deleted. See billing & spend controls.

Two bills — the deliberate split: PaaSbox bills the flat per-cluster fee for the control plane; Hetzner bills your worker nodes, load balancers, and volumes directly to you. Explained on the pricing page.

Under the hood

The technology operating all of this is Gardener, the open-source Kubernetes fleet manager from SAP — “Kubernetes that runs Kubernetes”, battle-tested on fleets of thousands of clusters. Gardener has its own vocabulary (you may encounter it in Kubernetes metadata on your cluster), which maps to the nouns above: a shoot is your cluster, a seed is the platform infrastructure your control plane runs on, and the garden is the platform’s own control plane. You never need these words to use PaaSbox — but if you see them in a label or an event, that’s what they mean. More in how paasbox works.