[SOLVED] XZ-compressed data is corrupt upon VM backup restore to updated Proxmox

[Ice Drake]

I am in the progress of migrating my LXCs and VMs from Proxmox VE 5.3-6 to 6.4-4. All the backups of LXCs transferred and restored works correctly as expected; however, my only VM transferred and restored didn't work as expected. The backup restore succeeds successfully, but upon booting it, I keep getting the same following error:

Loading ../vmlinuz-4.14-138-rancher...ok
Loading ../initrd-v1.5.6...ok


XZ-compressed data is corrupt

 -- System halted

I backup the VM twice while it is turned off via GUI and restore it multiple time from the two backups, but the result is the same. I am wondering if it has to do with how RancherOS is setup, but considering the backup works in Proxmox, it should be a straight forward process. It doesn't make sense why I am getting this strange error.

 

[Stefan_R]

Can you post your VM config after restore? ('qm config <vmid>')

 

[Ice Drake]

Here is VM config after the backup restore, readjusting more memory, and adding a CD-ROM:

root@pandora:~# qm config 202
boot: order=scsi0;scsi1;net0;net1;net2;net3
cores: 2
memory: 3072
name: Rancher
net0: virtio=1E:92:E5:91:08:60,bridge=vmbr2
net1: virtio=E2:0B:2D:01:16:21,bridge=vmbr3
net2: virtio=CE:4A:FE:98:86:18,bridge=vmbr4
net3: virtio=42:AF:D5:EA:AE:51,bridge=vmbr5
numa: 0
onboot: 1
ostype: l26
scsi0: local-lvm:vm-202-disk-0,size=10G
scsi1: local:iso/rancheros-autoformat_1_.iso,media=cdrom,size=153M
scsihw: virtio-scsi-pci
smbios1: uuid=e28216c7-8a89-45d9-b338-6e3c8878cade
sockets: 1
startup: up=60
vmgenid: 70470d52-22d4-4fd2-a1a3-1783386cb722

Here is VM config before the migration:

root@Pandora:~# qm config 202
bootdisk: scsi0
cores: 2
memory: 2048
name: Rancher
net0: virtio=1E:92:E5:91:08:60,bridge=vmbr2
net1: virtio=E2:0B:2D:01:16:21,bridge=vmbr3
net2: virtio=CE:4A:FE:98:86:18,bridge=vmbr4
net3: virtio=42:AF:D5:EA:AE:51,bridge=vmbr5
numa: 0
onboot: 1
ostype: l26
scsi0: local-lvm:vm-202-disk-0,size=10G
scsihw: virtio-scsi-pci
smbios1: uuid=e28216c7-8a89-45d9-b338-6e3c8878cade
sockets: 1
startup: up=60
vmgenid: 38ef1fb9-922e-46ca-9127-fa471da1ce5a

You can see that it is almost the same as before the migration except the part where I added the CDROM and have the system booted from it. Given that it is a system docker, I found that I am able to get everything to work correctly as expected if I booted from the CDROM as its system docker substitute instead of booting it directly from the disk drive. Thus, I suspected that maybe there is something different between the two Proxmox versions causing this strange behavior.

 

[Stefan_R]

Hm, you could try booting from a live restore image and checking the original drive from that, just to see if all the data is in fact correctly restored.

Also, compare the checksums (sha1sum /dev/mapper/pve-vm--202--disk--0) of the two disks before and after restore, if they differ, use something like 'vbindiff' to see where. The 'XZ-compressed data is corrupt' message would indicate that the BIOS and QEMU startup paths executed successfully, and the error only occured during loading of the actual image from disk.

 

[Ice Drake]

Yeah, I checked the checksum of the disk on the old Proxmox server before the restore and of the disk on the new Proxmox server after restore, and it doesn't match up.

root@Pandora:/dev/pve# echo --n vm-202-disk-0 | md5sum
06d05d82b8d1a09433d6721cc1ac79f6  -

root@pandora:/dev/pve# echo --n vm-202-disk-0 | md5sum
c93da46e6642afc91c68fbc59776221c  -

Since I don't remember how I setup that VM in the past and booting it up from a boot disk works fine, I am going to move on and keep it as is. It is taking too long to troubleshoot this issue. Thanks.