Is a single node enough?

For learning, staging and most early-production workloads, yes. Even the smallest size (2 vCPU / 4 GB, dedicated) comfortably runs a typical web app, its database and ingress for your first few thousand users. skube runs on skube: this very website is served from our own single-node cluster. And when multi-node HA starts paying off, your manifests move to an HA cluster unchanged.

What happens if the server fails?

Hetzner’s platform handles most hardware failures by restarting the VM with disks intact. If a cluster stays down, email us and we restore it from the nightly backup: getting it back is our top priority. A restore can lose up to a day of changes, so keep your own backups for critical data.

What happens when I outgrow it?

You migrate, and because it is stock Kubernetes, nothing about your workloads changes. The same manifests and GitOps that ran here apply cleanly to an HA cluster when you need multi-node or higher availability.

EKS & GKE alternative: a managed Kubernetes comparison

An alternative to EKS and GKE: a dedicated cluster we run for you on EU infrastructure, at one flat monthly price (€24, €59 or €109), on standard Kubernetes you can pick up and move. Here is how it compares.

Create a cluster

Spec-matched estimates: same RAM, same SSD, one load balancer. Method and sources under the table.

More info

	skube	EKS	GKE
skube12 vCPU · 4 GB · 40 GB	€24 flat	from ~€108 est.t3.medium · 2 vCPU · 4 GB	from ~€107 est.first cluster from ~€43e2-medium · 2 vCPU · 4 GB
skube24 vCPU · 8 GB · 80 GB	€59 flat	from ~€137 est.t3.large · 2 vCPU · 8 GB	from ~€135 est.first cluster from ~€71e2-standard-2 · 2 vCPU · 8 GB
skube38 vCPU · 16 GB · 160 GB	€109 flat	from ~€196 est.t3.xlarge · 4 vCPU · 16 GB	from ~€189 est.first cluster from ~€125e2-standard-4 · 4 vCPU · 16 GB
Pricing model	Flat	Control plane + nodes + egress, metered	Cluster fee + nodes + egress, metered
Traffic / egress	20 TB included	Metered per GB	Metered per GB
Disaster recovery	Nightly snapshots included	DIY or paid add-on	Paid add-on
High availability	No, one dedicated node	Multi-AZ control plane, node groups	Regional clusters, multi-zone
Autoscaling	No, fixed size per cluster	Node groups, Karpenter	Node pools, Autopilot
Setup to first kubeconfig	One order form	VPC, IAM, node groups, add-ons	Project, node pools, IAM
Who runs it	We do, end to end	Control plane managed; nodes, upgrades, add-ons on you	Control plane managed; node pools on you
Standard, upstream k8s	Yes, stock	Conformant, AWS add-ons	Conformant, GCP add-ons
EU data residency	EU only	EU regions	EU regions

How these estimates work, disclaimer & trademarks

Competitor estimates use the cheapest instance matching each tier's RAM (see More info for instance and vCPU), the same SSD, and one load balancer with minimal egress; publicly available prices as of 2026-07-06, excl. tax (skube prices shown include VAT), USD converted at €1 = $1.1424. Headline GKE figures exclude its cluster-fee credit; the "first cluster" line includes it (the credit covers one zonal cluster per account; the fee applies from the second cluster on). Competitor pricing and features change; check each provider's own site for current details.

EKS, GKE and Kubernetes are trademarks of their respective owners, used here for identification only; their owners do not endorse skube.

Where skube fits (and where it doesn't)

skube is one dedicated node. No failover, no autoscaling. For a web app, its database and an ingress serving a few thousand users, it is simply cheaper: the gap between €24 flat and the ~€108 estimates above is redundancy you were paying to not need. If you do need it, multi-AZ failover, autoscaling fleets, deep AWS or GCP integration, stay on EKS or GKE. They are priced for that job and they are good at it.

Because everything here is standard Kubernetes, moving between skube and the big clouds later is a kubectl apply, not a rewrite.

Create a cluster

Free account; pay only when you order a cluster. EU consumers have a 14-day right of withdrawal. Questions first? Email support@skube.dev, a human answers.

Frequently asked

Is a single node enough?: For learning, staging and most early-production workloads, yes. Even the smallest size (2 vCPU / 4 GB, dedicated) comfortably runs a typical web app, its database and ingress for your first few thousand users. skube runs on skube: this very website is served from our own single-node cluster. And when multi-node HA starts paying off, your manifests move to an HA cluster unchanged.
What happens if the server fails?: Hetzner’s platform handles most hardware failures by restarting the VM with disks intact. If a cluster stays down, email us and we restore it from the nightly backup: getting it back is our top priority. A restore can lose up to a day of changes, so keep your own backups for critical data.
What happens when I outgrow it?: You migrate, and because it is stock Kubernetes, nothing about your workloads changes. The same manifests and GitOps that ran here apply cleanly to an HA cluster when you need multi-node or higher availability.