A kernel update failed partway through on my encrypted Linux laptop. The boot partition was full, the update left the system in a half-configured state, and the machine wouldn't boot. What followed was four to five days of continuous diagnostic work — booting from live media, unlocking the encrypted volume, reclaiming space, repairing the boot configuration, and finishing the update. No data loss.

This post walks through what happened, what I tried, and what finally worked.

The Failure

The laptop was running an encrypted Linux installation. During a routine kernel update, the process failed partway through. The root cause was a full /boot partition — there wasn't enough space to install the new kernel files alongside the old ones, and the update died mid-write.

The result was an unbootable system. The old kernel's boot configuration had already been partially overwritten, and the new kernel hadn't finished installing. GRUB couldn't find a valid kernel to load. The machine dropped to a GRUB rescue prompt or failed to POST into anything useful.

The complication: the root filesystem was on an encrypted volume. You can't just mount it from a live environment without the passphrase and the right tools to unlock it first. Every repair step had to go through the unlock process.

Booting From Live Media

The first step was booting from a live USB. From there I could access the disk, but the encrypted volume was locked. I needed to identify the partition layout, find the encrypted partition, and unlock it before I could do anything with the actual filesystem.

I used the live environment's tools to open the encrypted volume with the passphrase, then mounted the root filesystem and the boot partition to a working mountpoint. This gave me read-write access to the broken system's files from outside the broken system.

Diagnosing the Partition Layout

The core problem was clear: /boot was full. Old kernel images, initramfs files, and leftover package data had accumulated over time and never been cleaned up. The failed update had tried to write new files into a partition that couldn't hold them.

I worked through the partition layout to understand exactly what was consuming space. The goal was to reclaim enough room on /boot to either complete the interrupted update or roll back to a known-good kernel.

This meant identifying which kernel versions were still referenced by GRUB, which ones were safe to remove, and which package files were left in a half-installed state by the interrupted update. I removed stale kernel images and their associated initramfs files to free up space.

Repairing the Boot Configuration

With space reclaimed, I needed to fix the boot configuration so the system could actually start. The interrupted update had left GRUB's configuration pointing at kernel files that either didn't exist or were incomplete.

From the live environment, I chroot'd into the mounted root filesystem. This let me run the system's own package manager and boot tools as if I were inside the installed OS. From the chroot, I could:

  • Clean up the interrupted package state so the package manager would function again
  • Complete the kernel installation that had been interrupted
  • Regenerate the initramfs for the working kernel
  • Rebuild the GRUB configuration to point at valid kernel and initramfs files

Each step had to be done in the right order. The package manager had to be functional before I could finish the kernel install, the kernel had to be installed before generating the initramfs, and the initramfs had to exist before GRUB could be configured to use it.

Completing the Update

After rebuilding the boot configuration, I unmounted everything, closed the encrypted volume, pulled the live USB, and rebooted. The system came up cleanly on the new kernel with all data intact.

The entire process took four to five days of continuous work. Not because any single step was impossibly hard, but because every step required research, testing, and careful execution. One wrong move on an encrypted volume with no backup of the encryption headers could have meant permanent data loss.

What I Took Away

This was one of the more involved recovery jobs I've done, and it reinforced a few things:

  • Boot partitions need monitoring. A full /boot is a ticking time bomb on any system that doesn't auto-clean old kernels. In a production environment, this is the kind of thing that should be caught by disk space alerts long before it causes a failure.
  • Encryption adds real complexity to recovery. Every diagnostic step requires unlocking the volume first. If you lose the passphrase or the LUKS header is damaged, the data is gone. Backups of encryption metadata matter.
  • Chroot is the single most useful recovery technique for Linux. Being able to step inside a broken system from a live environment and run its own tools is what makes most of these repairs possible at all.
  • Patience pays off. Five days is a long time to sit with a broken laptop. But the alternative was a reinstall and potential data loss. Taking it slow and understanding each layer before touching it is what made the difference.

If you're dealing with something similar — an encrypted system that won't boot after an interrupted update — the process is: live media, unlock the volume, mount everything in the right order, chroot in, fix the package state, rebuild the boot chain. The specifics depend on your distro and encryption setup, but the pattern is the same.

← Back to Blog