Wipe a disk before provisioning
The sealed-node provisioner refuses to install onto a disk that still holds RAID or a foreign operating system, and it does not wipe the disk for you. Clear it yourself in the Hetzner rescue system before you register the server. Wiping is irreversible.
What blocks an install#
Syself Autopilot refuses to install onto the target disk in any of these cases. Each one can mean the server is still in use, or that the target is not the disk you think it is, so Syself Autopilot stops rather than risk wiping the wrong data.
| Blocker | What it means |
|---|---|
| RAID | A disk in the server carries a RAID member signature or an assembled RAID array. Sealed nodes run on plain disks only. |
| Ambiguous WWN | More than one disk reports the target's WWN (World Wide Name, a disk's unique hardware ID), so the target is not uniquely identified. |
| LVM physical volume | The target holds an LVM physical volume, part of the Linux Logical Volume Manager. |
| Foreign Linux | Another disk in the same server holds a Linux installation. |
Warning
Wiping a disk destroys all data on it, permanently. Confirm you have the right server and that nothing on it needs keeping before you continue.
Why Syself does not wipe for you#
A wipe cannot be undone, and the platform never destroys data it did not create. A server you just bought can still carry a previous owner's array, and on a multi-disk server the disk you think is empty might be the one you meant to keep. Rather than guess, Syself Autopilot stops and asks you to clear the disk yourself. You do this once, when you first bring the server in, so a forgotten setting can never turn a data disk into a wiped one later. See Order and prepare a Robot server for where this fits when you bring a server online.
Clear the disk in the rescue system#
If the server is already a node, remove it from the cluster first.
Boot into rescue #
Open the Hetzner Robot console and select the server.
- On the Rescue tab, choose your SSH key and keyboard layout, then activate the rescue system.
- On the Reset tab, send a CTRL+ALT+DEL reset.
The server reboots into rescue after a few minutes.
Wipe the disks #
SSH in with the key you selected, then list the disks:
$ lsblk
Disk names look like nvme0n1 and nvme1n1, or sda for SATA devices. Wipe each one:
$ wipefs -af /dev/nvmeXnX # replace nvmeXnX with the disk name from lsblk
For a disk that is part of a software RAID array (managed by mdadm, Linux's software-RAID tool), stop the array first:
$ mdadm --stop /dev/mdX
$ wipefs -af /dev/nvmeXnX
Repeat for every disk that needs clearing.
Once the disks are clean, register the server (or return it to your cluster configuration). Syself Autopilot claims the HetznerBareMetalHost, partitions the target disk itself, and starts provisioning. You do not need to partition or format anything by hand: clearing the blocking signature is enough.
Related: add bare-metal servers , the serial console and rescue system , and a server that will not provision .
Remove a specific node
Pick exactly which node gets removed when you scale a pool down, instead of leaving the choice to Syself Autopilot.
Machine health checks and remediation
Syself Autopilot watches your nodes and replaces the ones that go unhealthy on their own, unless you pause it first for planned maintenance.