1 year PowerDedi turns 1: −20% on every server with code Ends in 14 h Claim offer

Pay with
PowerDedi
Deploy a server
Log in Deploy a server

Server management

Repair your server with rescue mode

Boot a Debian-based live system to mount your disks, fix boot problems, reset passwords or copy data off.

Updated September 25, 2026 · 3 min read · 7 sections

On this page 7 sections
  1. Enable rescue mode
  2. Log in to the rescue system
  3. Identify disks and arrays
  4. Mount the system and chroot into it
  5. Common repairs
  6. ZFS pools
  7. Boot from disk again

Rescue mode boots your server into a Debian-based live system loaded over the network, independent of the installed operating system. Use it to repair a server that doesn’t boot, reset a lost password or copy data off. You need access to the client area and, ideally, an SSH key saved under Account → SSH keys.

Enable rescue mode

  1. In the client area, open Servers, select the server and open the Rescue tab.
  2. Your saved SSH keys are installed in the rescue system. A temporary root password also appears on the tab once the system is ready: reveal it with one click; it is deleted after 24 hours.
  3. Click Boot into rescue: the server restarts into the rescue system, even if the installed system is frozen. Progress shows on the tab.

Rescue mode stays active until you switch the server back to booting from disk, as described at the end of this guide.

Log in to the rescue system

The rescue system has its own SSH host key, so your SSH client would warn that the host identification has changed. A separate known-hosts file keeps both keys apart:

ssh -o UserKnownHostsFile=~/.ssh/known_hosts_rescue root@203.0.113.10

The rescue system runs from memory: anything you change outside your mounted disks disappears at the next reboot.

Identify disks and arrays

lsblk -o NAME,SIZE,TYPE,FSTYPE,MODEL,SERIAL
cat /proc/mdstat

Device names can differ from those in your installed system, so recognize disks by size, model and serial number. Linux installs use mdadm RAID by default. If /proc/mdstat doesn’t list your arrays yet, assemble them:

mdadm --assemble --scan
cat /proc/mdstat

Arrays may get unusual numbers, such as md126 or md127; that is harmless. If your layout uses LVM, activate it with vgchange -ay and list the volumes with lvs.

Never run mdadm --create on existing disks to “repair” an array. It writes new metadata and can destroy your data. Use --assemble, and ask support if an array refuses to start.

Mount the system and chroot into it

Mount the root filesystem on /mnt, then read its /etc/fstab to see what else to mount. The device names below are examples:

mount /dev/md1 /mnt
cat /mnt/etc/fstab
mount /dev/md0 /mnt/boot             # only if /boot is separate
mount /dev/nvme0n1p1 /mnt/boot/efi   # only on UEFI systems

Bind the kernel filesystems into the mounted system and switch into it. From then on, commands run as if the installed system had booted:

mount --rbind /dev /mnt/dev && mount --make-rslave /mnt/dev
mount --rbind /sys /mnt/sys && mount --make-rslave /mnt/sys
mount --bind /run /mnt/run
mount -t proc proc /mnt/proc
chroot /mnt /bin/bash

Common repairs

Fix /etc/fstab

A wrong UUID or a missing disk in /etc/fstab stops the boot. Compare the file with the output of blkid, correct it, then check it with findmnt --verify. Add the nofail option to disks the system can boot without.

Reinstall the boot loader

Find out how the server boots with [ -d /sys/firmware/efi ] && echo UEFI || echo BIOS. On Debian and Ubuntu:

grub-install /dev/nvme0n1 && grub-install /dev/nvme1n1       # BIOS: both disks
grub-install --target=x86_64-efi --efi-directory=/boot/efi      # UEFI
update-grub

On AlmaLinux and Rocky Linux, run grub2-mkconfig -o /boot/grub2/grub.cfg, preceded on BIOS systems by grub2-install for each disk. On Proxmox VE installed on ZFS, run proxmox-boot-tool refresh instead.

Reset a password or restore SSH access

Run passwd root or passwd alice, or append your public key to /root/.ssh/authorized_keys. On AlmaLinux and Rocky Linux, also run touch /.autorelabel: SELinux then relabels the changed files at the next boot, which takes a few extra minutes. Without it, logins can fail.

ZFS pools

Proxmox VE and FreeBSD installed on a ZFS mirror use a ZFS pool instead of mdadm. List the pools available for import, then import yours under /mnt:

zpool import
zpool import -f -R /mnt rpool

Proxmox VE usually names its pool rpool; the FreeBSD installer uses zroot. Export the pool with zpool export rpool before you reboot, or the installed system may refuse to import it. A FreeBSD pool can use features that the rescue system’s ZFS doesn’t support. In that case, or if zpool isn’t available, boot the installer ISO of your system from the console and use its live shell.

Boot from disk again

exit                 # leave the chroot
umount -R /mnt
  1. In the Rescue tab, click Boot from disks.
  2. Reboot the server, for example with reboot from the rescue shell.
  3. Watch the boot on the Console tab. If it still stops, the screen shows where, which tells you what to fix next.

Related guides

All guides