LVM backup restoration fails under BRTFS pool

[tlaurion]

This is

sudo wyng version
Wyng 0.8 beta release 20240528

And wyng-util-qubes from 09beta.

---

Some trace extract:

sudo wyng-util-qubes --dest qubes-ssh://wyng-WRT3200ACM_raid5:root@Insurgo-WRT3200ACM/mnt/Backups/nv41 --authmin 1080 -w debug --debug restore qusal -u
	wyng-util-qubes v0.9 beta rel 20240530
	['/usr/sbin/wyng', '--dest=qubes-ssh://wyng-WRT3200ACM_raid5:root@Insurgo-WRT3200ACM/mnt/Backups/nv41', '-u', '--authmin=1080', '--debug', '--json', 'list']
	['/usr/sbin/wyng', '--dest=qubes-ssh://wyng-WRT3200ACM_raid5:root@Insurgo-WRT3200ACM/mnt/Backups/nv41', '-u', '--authmin=1080', '--debug', '-u', '--quiet', '--save-to=/tmp/wuqibythqw8/qmeta.tgz', '--session=20240531-163158', 'receive', 'wyng-qubes-metadata']
	['tar', '-xzf', '/tmp/wuqibythqw8/qmeta.tgz']

	VMs matched in session 20240531-163158:
	 qusal

	Restoring VM data volumes:
	Wyng 0.8 beta release 20240528
	Fetched archive.ini 5847
	Fetched archive.salt 424
	Fetched salt.bak 424
	metadata cipher = xchacha20-poly1305-msr
	[var]
	uuid = ce8a1804-a0f9-408f-94bf-989552462c48
	updated_at = 1717188583.8891625
	format_ver = 3
	chunksize = 131072
	compression = zstd
	compr_level = 7
	hashtype = hmac-sha256
	ci_mode = 35
	dataci_count = 4459606
	mci_count = 93475

alias vm-qusal-private = appvms/qusal/private.img
Local storage is offline: '/var/lib/qubes' is not a subvolume.
**fstype is btrfs
**pooltype rlnk not online
/var/lib/qubes None
Encrypted archive 'qubes-ssh://wyng-WRT3200ACM_raid5:root@Insurgo-WRT3200ACM/mnt/Backups/nv41' 
Last updated 2024-05-31 16:49:43.889163 (-04:00)

 Offline volumes: vm-qusal-private
JSON:
 {"/var/lib/qubes": [["vm-qusal-private", "appvms/qusal/private.img"]]}
['/usr/sbin/wyng', '--dest=qubes-ssh://wyng-WRT3200ACM_raid5:root@Insurgo-WRT3200ACM/mnt/Backups/nv41', '-u', '--authmin=1080', '--debug', '--force', '--sparse-write', '--local-from=/tmp/wuqibythqw8/wuq_vols.lst', '--session=20240531-163158', 'receive']
Traceback (most recent call last):
  File "/usr/sbin/wyng-util-qubes", line 409, in <module>
    handle_wyng_vol_error(p)
  File "/usr/sbin/wyng-util-qubes", line 162, in handle_wyng_vol_error
    errln = [x for x in text.splitlines() if x.startswith("Error on volumes:")][0]
            ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^
IndexError: list index out of range

[tasket]

@tlaurion What this looks like is that /var/lib/qubes was not setup as a subvolume, and the util mis-parsed the error message.

[tasket]

@tlaurion You are new to Btrfs so may want to look into how many pools the Qubes installer created with qvm-pool. Also check which one is default with qubes-prefs default_pool (you can use it to change the default). Possibly '/var/lib/qubes' is not where you want to be restoring VMs?

[tasket]

@tlaurion Let me know if you need guidance on converting the Qubes pool into a subvolume. I am going to relax this requirement for receive/restore, but it will still be necessary for send/backup unless someone thinks of an efficient workaround.

[tlaurion]

@tasket that testing laptop had bees installed before restoring. Maybe that changes the subvolume volume setup as opposed to the installer defaults for btrfs over q4.2.1? I will be able to test Tuesday on non-bees setup laptop replicate of bees setup, with only difference being bees deployed configured and running.

Self build rpm for fedora-37 rpm at QubesOS/qubes-issues#6476 (comment)

Discussion under https://forum.qubes-os.org/t/bees-and-brtfs-deduplication/20526

Qubes-builderv2 poc (non-working yet) at https://github.com/tlaurion/qubes-bees where spec file there was used to build rpm manually.

[tlaurion]

@tlaurion Let me know if you need guidance on converting the Qubes pool into a subvolume. I am going to relax this requirement for receive/restore, but it will still be necessary for send/backup unless someone thinks of an efficient workaround.

@tasket does wyng requires anything to be done on top of q4.2.1 default installer option other then choosing btrfs as partition scheme?

I will deferenciate bees modified btrfs subvolume options vs standard install of btrfs but yeah, I'm new to btrfs.

I just need to understand how to setup things to have things compatible with bees so that on receive/restore ops, bees dedup on the fly block volumes and if I understand well, bees only works on subvolume so I guess bees modified something and I missed the details because unaware.

[tasket]

@tlaurion Yes, whichever path is being used by the Qubes reflink pool (possibly /var/lib/qubes) needs to be converted into its own subvolume (instead of just being a plain dir on a Btrfs filesystem) for send and monitor to work. I think I posted a shell script to do that somewhere but I don't remember now. There is a good script in this Qubes forum post; the main thing I'd do different is to also stop the 'qubesd' service right after the qvm-shutdown step, then start it again at the end.

[tasket]

BTW, receive should no longer have any issue with restoring to a non-subvolume path or a path on a non-CoW storage system (so non-subvol, or Ext4 or whatever is now OK). Maybe I'll even get receive working with non-Linux systems.