Use a server you already have
Instead of a new server, a cluster can run on a server that is already in your Hetzner project. The portal rebuilds it with the node image: the server keeps its ID, its IP addresses and what Hetzner bills you for it. Its disk is erased.
Why you would
Section titled “Why you would”Hetzner raised its cloud prices for new orders on 15 June 2026. A server ordered before that date keeps its old price as long as it exists and is not rescaled. A rebuild keeps the server, so a cluster on it keeps the old price. The portal never rescales an adopted server and never deletes it, and an upgrade replaces the image in place without a rebuild. Whatever Hetzner bills you for the server today, it bills you after the adoption.
Which servers qualify
Section titled “Which servers qualify”The form lists every server of the project. Those that cannot be adopted are greyed out, each with its reason:
| Requirement | Why |
|---|---|
| A server type of the CPX, CCX or CAX line | The node image boots with UEFI only. CX servers boot with BIOS only and cannot start it. |
| An architecture the current release has an image for | The image is built per architecture. |
| A location the portal offers to your project | The cluster is created there. |
| Running or off, with no action running on it | The rebuild needs a server at rest. Wait for a running action to end. |
| Rebuild protection off | The portal cannot rebuild it otherwise. Hetzner turns delete protection off together with rebuild protection, so the server has neither afterwards. |
| A public IPv4 address | The cluster’s API name points at it. |
| Not part of another cluster or a node pool | One server, one owner: a server that carries the labels of another cluster or a node pool is refused. |
No private network that overlaps 10.0.0.0/16 | That is the cluster’s network. Detach the server from the overlapping network first. |
| No firewall that applies to it through a label selector | The portal can take a server out of a firewall it was added to directly, but not out of one that selects it by label. Change that firewall first. |
Adopt a server
Section titled “Adopt a server”-
On the create form (Create a cluster), choose your Hetzner project and, under Server, A server I already have.
-
Choose the server in the list. Its type, location and architecture become the cluster’s, and the list shows Hetzner’s list price for the type; a server you have had for a while may be billed at its older price.
-
Read what adopting it does. If volumes are attached, tick that they stay attached and are not erased; the cluster does not use them. If the server was added to firewalls directly, tick Take it out of my firewalls: the cluster’s firewall replaces them, with the API port from your ranges, 80 and 443, and ICMP. Otherwise take it out of them in the Hetzner console first.
-
Type the server’s name to confirm that its disk is erased. A rebuild cannot be undone.
-
Fill in the rest of the form as for a new server and choose Create cluster. The topology is Single node.
The portal then labels the server and renames it to the node’s name, <DNS label>-cp-1, attaches it to the cluster’s network, takes it out of the firewalls you chose, waits until the cluster’s firewall covers it, rebuilds it with the node image, and starts it if it was off. Its own primary IPv4 stays the cluster’s address; no new one is created. Once pbx-agent on the server has enrolled, the portal never rebuilds that server again, also not on a retry.
What happens to the server
Section titled “What happens to the server”| Kept | Changed | Not touched |
|---|---|---|
| Its ID | Its disk: rebuilt with the node image; everything on it is gone | The volumes attached to it |
| Its IP addresses; the API’s name points at its IPv4 | Its name: the node’s name | Its Hetzner snapshots and backups |
| What Hetzner bills you for it | It joins the cluster’s network and firewall, and carries the cluster’s labels | Your other labels on it, and the private networks it is already in |
When you delete the cluster
Section titled “When you delete the cluster”The server is never deleted. The portal hands it back: it leaves the cluster’s network, loses the cluster’s labels and with them the cluster’s firewall, and gets its old name back. It keeps running as it is, with the cluster’s system and data still on its disk and no firewall of the portal’s in front of it. Stopping, rebuilding or deleting it is up to you, and Hetzner keeps billing it to you. In the lab the delete took 23 seconds. The rest of the cluster’s resources are deleted as usual (Detach or delete a cluster).