Lost Partition Tables on VMs

[mliebherr99]

Hello,

I am running:
pve-manager/4.4-22/2728f613 (running kernel: 4.4.117-1-pve) => node01
pve-manager/4.4-13/7ea56165 (running kernel: 4.4.44-1-pve) => node02
pve-manager/4.4-22/2728f613 (running kernel: 4.4.117-1-pve) => node03

and i use ceph/rbd as disk.

After my node02 crashed. A few VMs lost their partition table and i had to restore them with testdisk

I use "Default (No cache)".

Any idea how that happened? Did that happen to anyone else before?

Thanks,
Mario

 

[Jan Sievertsen]

Hi,
older thread but i run into similar problems. From time to time one Vm loses their partition table. Any ideas for help would be great.

 

[przemeksw]

I have a similar problem. I suspect a discard option. On several masts where I have it turned off, the problem does not occur.

 

[CodeNinja]

I noticed the same after i upgraded from 6.x to 7.x
It seems that they are already lost longer as the backups and snapshots also does not contain the partition table.
Now i have 10 broken machines. I could fix 8 of them with test disk, the other 2 are lost.
This is NOT ACCEPTABLE!

It seems that more people have similar issues. Whats the cause?

 

[RolandK]

i guess it's a long hidden bug in kvm/qemu, maybe related to qemu. there is a bugticket at https://bugzilla.proxmox.com/show_bug.cgi?id=2874

 

[sigmarb]

+1 hit me today. Guest was Debian 10 system.

 

[d1_sen]

+1, I'm still trying to recover my centos guest system, no luck so far

 

[RolandK]

>I have a similar problem. I suspect a discard option. On several masts where I have it turned off, the problem does not occur.

so, does this apply for your VMs ?

 

[d1_sen]

no, the hd of the VM that I'm having issues with had the discard option disabled. were you able to resolve your issue using testdisk?

 

[RolandK]

i don't have the issue (at least i'm happy i was not hit yet), but as we currently move a larger env from xenserver to proxmox, i really want this getting resolved, as this is a real bummer for me. it makes me nervous. enterprise products should not have such bug. loosing disk partitions from out of nowhere must be resolved.

so i'm trying to help getting a clearer picture of the problem.

 

[d1_sen]

we have more than 100 VMs running on Proxmox, this is the first time we've encountered such an issue. But this is alarming because we have not kept separate backups of these VMs other than the Proxmox backup system. I dont understand why the backups are failing to restore properly.

 

[cetjunior]

Hi there,

+1 with this issue. Lost partition table of a legacy system, Debian 6, after the 1st backup of PBS. Recovered it with testdisk (over Clonezilla live). I've checked the backup status and it seems that it ended the backup, but PVE lost itself somehow when releasing the VM of the task. Will make some more tests.

EDIT: just adding: this specific system doesn't have guest tools (so, no discard option), it's a internal server but in production. Versions are PVE 7.1-10 and PBS 2.1-5.

Best regards.

 

[mic]

hi,

+1 with the issue. we migrated a uefi centos 7 from physical to proxmox via clonezilla.
it is now the second time that the vm lost the partition tables. we were able to recover the partition tables with gdisk.

we use pbs 2.2-5

br

 

[RolandK]

could you please provide information on what types of virtual disks have been used with the affected virtual machines?

 

[RolandK]

the bug (see bugticket above) was found and most likeley fixed today.