built in backup client restores mangled data. Guest doesn't start.

Dec 18, 2024
10
2
3
I have backups configured for all my guests in proxmox. Now for the first time I had to restore a linux server. ubuntu 24.04 thin LVM 256GB on disk. Restore task finishes OK. Start server. No disks/partitions. Why?

When start the guest it says it cant find the uuid and is dropping to abusybox shell in initramfs.

How do I recover a broken guest?
 
Last edited:
What is it now ... no disks/partitions or cannot find UUID? The later means that the disk was restored and now you have an error in your guest.

Best to boot a live linux and inspect what was restored, what the UUIDs are and what it expects.

Please provide also the current PVE version installed.
 
  • Like
Reactions: gfngfn256
It can't find the specified uuid and ubuntu drops in to a busybox in initramfs. PVE version 8.4.1 I Need no solution on restoring the guest. I need to know why something trivial as backup/restore doesn't work for me. During the recovery phase I had to fix some partition errors. Steps taken to get access to the data

The system has:

/dev/sda: 64GB SSD (system disk) (bare linux that has the guest drive attached)

/dev/sdb: 256GB secondary disk (This is the drive I am trying to restore)

/dev/sdb shows a corrupt primary GPT, but a valid backup GPT exists.

Two partitions detected on /dev/sdb:

/dev/sdb1: BIOS boot (1MB)

/dev/sdb2: Linux filesystem (256GB)

Initial Mount Attempts Failed

Mounting /dev/sdb or /dev/sdb2 failed due to:

Missing device node for /dev/sdb2

Possibly bad superblock or unrecognized file system

Package Installations to recover data

Installed file, parted, and confirmed kpartx was already installed.

Recovery Steps

Used kpartx -av /dev/sdb to map GPT partitions:

Created /dev/mapper/sdb1 and /dev/mapper/sdb2

Successful Mount

Mounted /dev/mapper/sdb2 to /mnt/guest

Browsed contents successfully; includes typical Linux filesystem structure

Recovered Data

Navigated to /mnt/guest/home/user/

Found directories and files like:

apps/, guest/, updates.sh, whisper.cpp, piperdata/, webui.db, etc.

Final Result

The partition was successfully mounted using kpartx after detecting a corrupt GPT.

All intended user data under /home/user/ was accessible and appears intact.
 
I need to know why something trivial as backup/restore doesn't work for me.
I get that, but without any more information, this is hard or almost impossible to guess.

During the recovery phase I had to fix some partition errors.
Oh good, so no Linux noob. For further investigation, I suggest you do a souple of backup & restore rounds in order to see if you can replicate the problem.
There were some reports in the past about GPT corruption on the forums, yet I do not know if they relate to this problem or if they're still current problems.
 
  • Like
Reactions: leesteken
And now what? I would really like some pointers as to why my restores fail to start.
What version of Proxmox VE are you running? What backup software are you using (PBS or built-in or other)? What storage are you using (for VM/CTs and backup)? Maybe you have drive or memory corruption (did you run a memtest and long SMART test)? What hardware are you using (ECC for example)? Did you take the suggestion of running a new backup and restore on another VM/CT for testing? Maybe share the VM configuration file (before and) after a failed restore? Did you maybe exclude virtual disks from backups? Please provide some concrete information and file/configuration contents.