Failed mounting pool after upgrade 5.21 to 6

Project Discussion LXD Support
1

Hello,
After changed the snap channel from 5.21/stable to 6/stable, the storage pool failed to mount.

Ubuntu 24.04.4
ssd ext4 for system + 1 ssd btrfs for LXD.

I tried sudo lxd recover:

This LXD server currently has the following storage pools:

- Pool “data” using driver “btrfs”

Would you like to continue with scanning for lost volumes? (yes/no) [default=yes]:

Scanning for unknown volumes… 

Error: Failed validation request: Failed mounting pool “data”: Failed to mount “/dev/nvme0n1” on “/var/snap/lxd/common/lxd/storage-pools/data” using “btrfs”: invalid argument

lxc storage show lxd-storage-pool:

Error: Storage pool not found

Thanks

2

Could you please share the output of: sudo blkid /dev/nvme0n1 and sudo fdisk -l /dev/nvme0n1?

3

Thanks for helping me.

It seems you’ve hit the nail on the head. Before, /dev/nvme0n1 was the ‘LXD’ disk, now it’s the system disk.

sudo blkid /dev/nvme0n1
/dev/nvme0n1: PTUUID=“8c215351-ed6b-4a19-9b09-eb06336f7a8a” PTTYPE=“gpt”

sudo fdisk -l /dev/nvme0n1
Disque /dev/nvme0n1 : 931,51 GiB, 1000204886016 octets, 1953525168 secteursModèle de disque : Samsung SSD 990 EVO Plus 1TBUnités : secteur de 1 × 512 = 512 octetsTaille de secteur (logique / physique) : 512 octets / 512 octetstaille d’E/S (minimale / optimale) : 512 octets / 512 octetsType d’étiquette de disque : gptIdentifiant de disque : 8C215351-ED6B-4A19-9B09-EB06336F7A8A

Périphérique     Début        Fin   Secteurs Taille Type/dev/nvme0n1p1    2048    2203647    2201600     1G Système EFI/dev/nvme0n1p2 2203648    6397951    4194304     2G Système de fichiers Linux/dev/nvme0n1p3 6397952 1953521663 1947123712 928,5G Système de fichiers Linux

sudo fdisk -l /dev/nvme1n1
Disque /dev/nvme1n1 : 1,82 TiB, 2000398934016 octets, 3907029168 secteurs
Modèle de disque : Samsung SSD 990 EVO Plus 2TB
Unités : secteur de 1 × 512 = 512 octets
Taille de secteur (logique / physique) : 512 octets / 512 octets
taille d’E/S (minimale / optimale) : 512 octets / 512 octets

4

That’s one of the reason why it’s recommended to use ID/UUID/label/etc (see /dev/disk/by-*/) instead of bare /dev paths as those are “enumeration order” dependent.

Now to get your setup back to functioning, I wonder if you can not just fix the source value with a lxc storage edit?

5

I can’t save the new config:

Config parsing error: Pool source cannot be changed when not in pending state

I have no container/vm in this pool (I have deleted the test container just before upgrade to 6.7). Maybe it’s easier to redo an init (if possible) ?

6

If you have no data to preserve, indeed, doing another init is likely the path of least resistance.

7

edit :

I used /dev/disk/by-id/ to create the storage.

Then change the default pool (with the UI)

Then delete the old pool (It’s a bit scary because it doesn’t say that it doesn’t wipe the data. And the pool was my system disk :slight_smile: )

Thank you very much for the help !

Sorry but I don’t see how to create a storage pool with a disk uuid.

If the disk is formatted:

ERROR: /dev/disk/by-uuid/4e4b850c-911d-4d46-b2cc-e1a152689ac8 appears to contain an existing filesystem (btrfs)

(response from init and from storage create pool)

If the disk is unformatted, the uuid doesn’t exist.

8

Indeed, suggesting the UUID approach was wrong, you are better off with /dev/disk/by-id/ like you did :slight_smile:

I’m not entirely clear if you’ve been able to fix the issue or not but just in case…

To clear the data: sudo blkdiscard /dev/.... This ideal as it lets the NVMe do some garbage collection while removing the data. If not supported by your device, sudo wipefs -a should work to remove the traces of the old btrfs FS and let LXD put a fresh filesystem on top.

9

You could use lxd sql global command to update the storage_pools_config table for the specific pool to the new location of the block device of the existing pool, rather than wiping and reinstalling.

1 Like