Kubernetes on Hetzner

The managed Kubernetes
Hetzner never shipped.

Hetzner sells the cheapest serious compute in Europe, and no managed Kubernetes to run on it. Ankra fills the gap: production clusters provisioned, operated, and repaired on your own Hetzner account.

€65/mo
24 vCPU production baseline
1 command
Full cluster topology
30 vCPU
Free allowance

Hetzner gives you servers. Production needs more.

The price per vCPU is unbeatable. The distance between a fresh VM and a production cluster is where teams lose weeks.

No managed control plane

There is no EKS equivalent. Hetzner hands you servers and a network; etcd, upgrades, and high availability are entirely your problem.

DIY glue everywhere

Terraform plus k3s scripts work beautifully until the person who wrote them leaves. Then your production cluster is a haunted house.

The invisible platform layer

Ingress, certificates, monitoring, logging, secrets. Fifteen Helm charts before your first deploy, each with a hundred settings and three that are load-bearing.

The 3am question

When a node dies at night, Hetzner's responsibility ends at the server. Everything above it belongs to whoever is awake.

The Ankra approach

One command.
The whole topology.

Ankra provisions managed k3s clusters directly on your Hetzner account: private network, bastion, control planes, worker node groups, optional ingress stack. Pick a location and instance types; the topology comes up on its own.

Everything on top, ingress, cert-manager, monitoring, your applications, lives in Stacks: versioned in Git, deployed as drafts you approve, reconciled by Ankra's GitOps engine.

ankra cli
ankra cluster hetzner create \
  --name prod-fsn1 \
  --location fsn1 \
  --control-plane-count 1 \
  --control-plane-server-type cx33 \
  --worker-count 3 \
  --worker-server-type cx43
network, bastion, control plane, workers: provisioning

One-command clusters

ankra cluster hetzner create brings up the private network, bastion, control planes, and workers on your own Hetzner account. A few clicks in the UI does the same.

Stacks, not fifteen Helm installs

Ingress, cert-manager, monitoring, and your applications live in versioned Stacks with dependency ordering. Every change is a draft you approve, committed to Git.

AI operations

A failing pod gets an incident report, not a page: the AI correlates logs, events, and deploy history, finds the cause, and drafts the fix as a commit.

Clone to any provider

The same stack lands on OVHcloud or UpCloud in an afternoon, variables resolved per cluster. A repricing becomes a decision, not a hostage situation.

Scale to zero

Idle staging clusters tear down to €0 in Hetzner charges and come back later with their stacks reconciled onto fresh servers. Pay for the hours you use.

Costs you can actually see

The Cloud Cost view estimates spend per cluster, namespace, and stack from live inventory and public pricing. The invoice stops being a monthly surprise.

Run the numbers

Two invoices, one workload

Drag the slider to your cluster size. Prices are current list prices; the caveats are printed right under the result.

Hetzner + Ankra vs AWS EKS

Same worker vCPUs, two invoices

Worker vCPUs
vCPU
4100200+
Hetzner + Ankra
3 × CX43 workers (8 vCPU / 16 GB)€47.97
Control plane (CX33)€8.49
Bastion host + IPv4€8.00
Ankra platform fee (inside the 30 vCPU free allowance)€0
per month€64.46
AWS EKS
EKS control plane$73
6 × m6a.xlarge workers (4 vCPU / 16 GB)$756
NAT gateways, 3 availability zones$98
Application load balancer$22
per month$949

Same worker vCPUs, roughly

13.6x cheaper

List prices, August 2026: Hetzner Germany/Finland monthly caps excluding VAT, EKS on-demand in us-east-1 before storage, egress, and observability. The multiple assumes $1.08 per €1. Shared Hetzner cores versus dedicated EC2 cores, one region versus three availability zones: this compares invoices, not SLAs. Savings plans and spot pricing narrow the gap, at the price of commitment or interruptions.

EKS, DIY, or Ankra

Three honest ways to run Kubernetes when the budget matters.

AWS EKSDIY on HetznerHetzner + Ankra
Monthly invoice, 24 worker vCPUs~$950 before storage and egress~€65 in servers, plus your evenings~€65 all in
Control plane$73/mo per clusterYours to babysitIncluded, never billed
Getting to a clusterAn afternoon of VPC and IAM wiringDays of Terraform and scriptsOne command or a few clicks
Ingress and certificatesAssemble it yourselfAssemble it yourselfOptional ingress stack included
Monitoring and loggingCloudWatch, billed per glanceRoll your own PrometheusStack add-ons, one deploy
When it breaks at 3amSupport tiers and ticket queuesYou, with kubectl and coffeeAI incident report with a drafted fix
Leaving laterDeep lock-inFree to leave, alone on arrivalClone the stack to another provider
Zero lock-in, on purpose

Cheap today. Portable when it stops being cheap.

Hetzner repriced in June 2026, and it will not be the last provider to do so. Because your platform layer lives in Stacks, the same setup clones to OVHcloud or UpCloud in an afternoon. Benchmark both with real traffic, read the Cloud Cost page, keep whichever invoice wins.

We worked the repricing math inHetzner Repriced: Here Is the Mathand benchmarked the providers head to head inSix Clouds, One Kubernetes Benchmark.

Questions teams actually ask

Does Hetzner offer managed Kubernetes?+

No. Hetzner sells excellent compute, networking, and load balancers, but has no managed Kubernetes product. Ankra provides that layer on top: it provisions and operates k3s clusters on your own Hetzner account, so the servers stay yours and the compute invoice stays Hetzner's.

What does production Kubernetes on Hetzner actually cost?+

A 24 worker vCPU baseline lands around €65 a month all in: three CX43 workers, a CX33 control plane, a bastion host, and IPv4 addresses at post June 2026 list prices, with the Ankra fee at €0 inside the 30 vCPU free allowance. The same footprint on AWS EKS is roughly $950 a month before storage and egress.

Is k3s production ready?+

Yes. k3s is a certified Kubernetes distribution maintained under the CNCF, exposing the same API your manifests already target. Ankra provisions it with a private network, a bastion, and your choice of control plane count and instance types.

What happened to Hetzner prices in June 2026?+

On June 15, 2026 Hetzner raised cloud prices: shared CX and Arm CAX lines went up roughly a third, while CPX and dedicated CCX families doubled or more. Existing servers keep their old price until resized. Even after the increase, Hetzner remains by a comfortable margin the cheapest serious place to run Kubernetes in Europe.

What if I want to leave Hetzner later?+

That is the point of the design. Your platform layer lives in Stacks that clone to OVHcloud or UpCloud in an afternoon, with variables resolved per cluster and secrets reconfigured per environment. Providers compete for portable workloads and tax captive ones.

How does Ankra pricing work on top of Hetzner?+

Ankra bills on worker vCPUs only. Control planes are free, every plan includes a 30 vCPU free allowance, and clusters and team members are unlimited. A small production cluster typically runs entirely inside the free tier.

Free tier available

Your first Hetzner cluster
fits in the free allowance

30 worker vCPUs included, control planes never billed, unlimited clusters and teammates. Bring your Hetzner account and a prompt.

Ankra is an independent platform and is not affiliated with or endorsed by Hetzner Online GmbH. Hetzner and AWS prices on this page are list prices as of August 2026 and can change; the linked pricing post tracks the details.