btrfs data rescue after I deleted a parent snapshot of my rollback
Can you expand a bit on that? I thought it didn’t matter if you deleted parent snapshots because the extents required by the child would still be there.
Honestly, I have no idea why it went wrong or why it let me do that. Also my memory is a bit fuzzy since it’s been a while, but as best I can remember what I did step by step:
fuck around with power management configs
using btrfs-assistant gui app, rolled back to before that
btrfs-assistant created an additional snapshot, called backup something, I didn’t really pay attention
reboot, all seemed good
used btrfs-list to take a look, the subvolume that was the current root / was a child of the aformentioned backup subvolume
started btrfs-assistant and deleted the backup subvolume
system suddenly read only
reboot, still read only
btrfs check said broken refs and some other errors,
i tried to let btrfs check fix the errors, which made it worse, now I couldn’t even mount the drive anymore because btrfs was completely borked
used btrfs rescue, which got all files out onto an external drive successfully
installed arch again and rsync the rescued files over the new install, everything works as before, all files are there
btrfs check said broken refs and some other errors,
Gotcha. That must have been a kernel bug (or hardware error), none of the userspace utilities could cause it unless they were trying to manipulate the block device directly, which would be really dumb. It’s possible it wasn’t even related to the subvolume manipulation.
Can you expand a bit on that? I thought it didn’t matter if you deleted parent snapshots because the extents required by the child would still be there.
Honestly, I have no idea why it went wrong or why it let me do that. Also my memory is a bit fuzzy since it’s been a while, but as best I can remember what I did step by step:
Gotcha. That must have been a kernel bug (or hardware error), none of the userspace utilities could cause it unless they were trying to manipulate the block device directly, which would be really dumb. It’s possible it wasn’t even related to the subvolume manipulation.