Some fatal backup errors cause the backup system to silently hang forever (unless backing up to dom0)

[rustybird]

Qubes OS release

Qubes OS 4.2

Brief summary

Some fatal errors during a (CLI or GUI) backup cause the backup system to suppress the relevant exception or error message and it just silently hangs forever, e.g.:

• Qubes Backup hangs and never finishes if any of the target qubes are running #7411
• Qubes Backup hangs if there is an I/O error #7567
• External USB drives possibly interrupting backup by going to sleep #7797 (comment)

Steps to reproduce

Using the deprecated 'file' driver as a convenient way to reproduce the problem:

qvm-pool add hang-pool file -o dir_path=/var/tmp/hang-pool
qvm-create -l red -P hang-pool hang-vm
qvm-create -l red dest-vm
qvm-start hang-vm dest-vm
qvm-backup --yes --passphrase-file=<(echo foo) -d dest-vm /var/tmp hang-vm

Expected behavior

qvm-backup quickly fails with the message Backup error: file pool cannot export running volumes, like it does when backing up to dom0 (without -d dest-vm).

Actual behavior

It hangs forever at Making a backup... 0.17%. No errors in the dom0 or dest-vm journal.

[rustybird]

@SarneWeber It would be interesting to know if backing up to dom0 also turns your silent hang from #7797 (comment) into a proper error. If you want to try this, you might still have to narrow down the list of VMs to be backed up so you don't have to cancel early (before the backup file fills up all the free space on your dom0 filesystem)

[SarneWeber]

I figured out what causes this issue; backing up my 0 bytes qube caused the backup to get stuck and never finish. In retrospect, I should have done a binary search sooner. Backing up the same 0 bytes qube to dom0 opposed to any other qube caused the error ERROR: Failed to find logical volume "qubes_dom0/vm-[name]-private". The error could be caused by any qube that has never booted before, or it could be caused by any qube that does not have any storage used. I'm unsure which of the two is the case. I suggest the following things need to be improved:

• any qube which does not have a logical volume is excluded from the gui, or grayed out instead. The cli should not attempt to backup these qubes either, but instead should send a message or warning that it skips these qubes. Possibly only in verbose mode.
• the error handling when backing up to any non-admin vm gets improved in case this somehow still happens

[rustybird]

Thanks! Looks like backing up to dom0 can indeed reveal the underlying error whenever backing up to a VM just hangs.

Backing up the same 0 bytes qube to dom0 opposed to any other qube caused the error ERROR: Failed to find logical volume "qubes_dom0/vm-[name]-private". The error could be caused by any qube that has never booted before, or it could be caused by any qube that does not have any storage used.

Can you open a new issue specifically for that bug? (I haven't been able to reproduce it)