Upgrading Raspbian Pi from Bullseye (11) to Bookworm (12)

I run a couple of Raspberry Pis at home for various tasks — DNS servers (Unbound), Pi-hole, and similar utilities. They’ve been running stably for over two years, but they’ve remained on Bullseye for quite some time.

I generally prefer running one OS version behind the latest to avoid late-breaking fixes or regressions. Another advantage of staying a version behind is that I mostly receive security patches rather than potentially unstable new package updates. My hardware is also a bit dated — one is a Raspberry Pi 3 and the other a Pi Zero W — so I tend to be a bit cautious that way!

Still, it was finally time for an OS upgrade.

Most guides online suggested flashing a new SD card with the latest OS and spending a couple of hours reinstalling everything. I wasn’t keen on that approach — surely there had to be a proper way to upgrade Debian in place. Both Pis were headless and quietly doing their jobs, so I really didn’t want to attach a monitor or keyboard just for this. I was looking for the laziest possible way to perform the upgrade remotely.

sudo apt update && sudo apt full-upgrade

Step 2: Update sources.list

Now came the acid test: updating the sources.list to point to Bookworm.

In youir After that, I ran:

Preparing to unpack .../ncurses-term_6.4-4_all.deb ...
Unpacking ncurses-term (6.4-4) over (6.2+20201114-2+deb11u2) ...
dpkg: error processing archive /var/cache/apt/archives/ncurses-term_6.4-4_all.deb (--unpack):
 unable to securely remove '/usr/share/terminfo/a/aaa.dpkg-new': Bad message
Errors were encountered while processing:
 /var/cache/apt/archives/ncurses-term_6.4-4_all.deb
E: Sub-process /usr/bin/dpkg returned an error code (1)

So I tried this

pi@raspberrypi:~ $ sudo apt clean

Followed by an attempt to fix any inconsistencies in the dpkg database.

pi@raspberrypi:~ $ sudo dpkg --configure -a

Then I attempted to fix my broken install

 sudo apt --fix-broken install

Nope that didn’t cut it. Then I noticed that the real issue was the filesystem itself. The directory was “unreadable” as far as apt was concerned. So I went into the directory to check the status myself as root. Sure enough the file system was corrupted. So time to run fsck

I realised this has to be run during reboot. Unfortunately with systemd and the vast knowledge of stackoverflow said I have to modify the kernel startup parameter as such.

I tried to simply reboot hoping somewhere in there it would recognise a corrupt filesystem and reboot.

So I did sudo reboot

And Lo! And behold, it booted back up and file system is fixed.

Restarted the full upgrade again. It finished without any issues this time

Then I did the cleanup routine

sudo apt autoremove --purge -y
sudo apt clean
sudo reboot

And first thing I checked

pi@raspberrypi:~ $ lsb_release -a
No LSB modules are available.
Distributor ID:	Raspbian
Description:	Raspbian GNU/Linux 12 (bookworm)
Release:	12
Codename:	bookworm

pi@raspberrypi:~ $ uname -a
Linux raspberrypi 6.1.21+ #1642 Mon Apr  3 17:19:14 BST 2023 armv6l GNU/Linux

Yes. It is done

Now the internet advised me to check on all services and their status

pi@raspberrypi:~ $ sudo systemctl --failed
  UNIT LOAD ACTIVE SUB DESCRIPTION
0 loaded units listed.

That seemed OK. Looks like the upgrade is a success! That was a major upgrade to a headless tiny box over the network! That is mighty impressive that Debian can handle such an upgrade without too many issues. Of course the disk corruption was a surprise that I didn’t anticipate.

My guess is everyone’s mileage may vary based on what they have installed and running. Mine was a simple PiHole and Unbound stock install with configuration that came through fine.