Skip to content

ssd just lost its partition table

So, I've just done a regular arch linux system update. Of course without kernel updates because of issues/14622.

After rebooting, there was nothing left. I still got an ZFSBootMenu entry somewhere in the efibootmgr but that's it. After using my archzfs usb iso, I went to software/arch-linux-configuration/scripts/zfs/recover and started 01-mount.sh. What I got in return was a no pool available.

I've used them all, parted, gdisk, testdisk and fdisk but they all have told me the same "shiny new ssd you got there". By the way, the ssd is a nvme-one from Samsung. I've fired up the lenovo internal system check tools but all they say is "super fine disk you got there, almost totally new".

What I did at the end was rebuilding my partition table using sgdisk. After that, I've also mkfs.vfat the efi boot partition and recreated the zfsbootmenu as well as the efibootmg entries. Important not, you have to update your /etc/fstab for the boot efi partition after doing that. Nice to know, after creating the zfs partition via sgdisk, I was able to fully import my encrypted zfs pool.

This is a pretty strange error and I hope I don't have to fix this again in the near future. It leaves a strange feeling in your stomach

Lessons learned? sgdisk -b=sgdisk-sda.bin /dev/sda ref

zfs-snap-manager and "Got invalid schema for dataset" or not deleted old snapshots or not created new snapshots

I am using the zfs-snap-manager for such a long time that I never ever had a look on it since years.

I've just created zfs pools and zfs snap manager configurations without thinking or checking it for a while now. But once upon a time I wanted to check the behaviour and found out that some configured snapshots where not made.

After a while I saw that my assumption was false, so the error was in front of the display :-).

My configuration for the section schema was simplified to 7d because all I wanted for a dataset was that I can go back the last seven days. After finding the log file (/var/log/zfs-snap-manager.log) and tailing it, I quickly saw the log message Got invalid schema for dataset .... Just a bit more investigation and I found out that I have to provide a fully defined schema like 7d0w0m0y. After adapting all available configuration files on all hosts, the zfs-snap-manager is working as expected.

I found an existing feature request to "relaxed" the schema interpretion and stumpled over another feature request asking for creating unittests. So I've forked it and will try to learn python by fixing the issue on my own.

  • migrate code to python3
  • write unittests
  • implement relaxed handling of schema with default values

Do you want to join? Contact me if you want to.

archzfs has a new key and a new maintainer for the unofficial arch linux user repository

The unofficial but great archzfs build has a new maintainer and most important, a new key.

#pacman-key -r F75D9D76
#pacman-key --lsign-key F75D9D76

And you are ready to upgrade. The key was changed since they are now using a buildbot to reduce the time between new kernel release and adapted archzfs package. Thanks for your great work people!