Emergency Backup Recovery without Qubes (v4): fails to untar 100+GB Qubes consistently

Hi,

I did a couple of disaster recovery simulations. Result is rather devastating.

Procedure:

  1. Install Qubes-OS 4.0.3

  2. Create Qubes with real data (small and big). Mine are from under a gigabyte up to over 130GB, including a standalone Qubes-Builder Qube.

  3. Work with it a week and do daily backups. In my case around 250GB, if all Qubes are selected.

  4. Attach the backup medium (in my case a brand new 4TB USB3 disk).

  5. Follow the documentation at Emergency Backup Recovery without Qubes (v4). BTW: I scripted step 6 to 8. I was to lazy (and disappointed) to write a full script.

  6. The biggest Qube (~130GB) fails consistently to untar:

    tar: Unexpected EOF in archive
    tar: Unexpected EOF in archive
    tar: Error is not recoverable: exiting now

  7. Delete private.img, and try again. In hope that it was just a hiccup, but fails too.

  8. In Qubes-OS backup only the Qube in question, and try again to restore. Same result.

  9. Since all of the above takes ages, I give up and rsync my data out of the live Qube in question. That will run probably until late in the night.

Note:
I started on Monday; today is Wednesday and I still haven’t recovered all the data. I used to work in enterprise disaster / recovery. If I would have needed more than one day to recover a storage system, I probably would have been fired the same day.

Conclusion:
The Qubes-OS backup is unreliable and in case of disaster by far to tedious, hence rather useless for production.

Below the script I used. Just to prove that I followed the documented procedure.

unpack-vm-image.cpiogz (646 Bytes)

Thank you for the bug report. I’ve opened an issue for it:

In the future, please feel free to file bug reports directly in qubes-issues, which allows us to handle them more efficiently.

Thank you for the bug report. I’ve opened an issue for it:

https://github.com/QubesOS/qubes-issues/issues/6113

In the future, please feel free to file bug reports directly in
qubes-issues, which allows us to handle them more efficiently.

Yep, you’re right I should have, but I was too busy too to get the data
out of Qubes-OS; and still am. I used rsync and copied the data old
school out on an external medium. So, now I’ve to mount all the other
imagines and rsync them into my traditional system (Manjaro Xfce).

Only a note: If that were a real disaster recovery (disk and/or disk
controller died) I would have lost around 130GB data and a week of
work. Currently I’m still good. A) I put time aside to test Qubes OS;
B) I have always at least 2 independent offline-backups; C) Disaster
Recovery is almost second nature to me.

However, thanks for filing this as bug. Disaster Recovery is an
important function.