[SOLVED] PVE 7.0: All VMs with Cloud-init & SeaBIOS fail during boot process (bootloop/disk not found)

[YAGA]

For several weeks now I have been stuck on this problem. All my old VMs created with Qemu / Cloud-Init since PVE 6.x no longer boot in PVE 7.x.

At the beginning they worked perfectly in PVE 7.x but an update (PVE update from community repo) carried out at the beginning of September now prevents the VM to boot in SeaBIOS. They are stuck in a boot loop in SeaBIOS.

Any ideas or suggestions welcome,

Regards,

 

[vmware]

I've given up waiting and slowly re-creating my VMs in PVE 7. It's a shame that this issue seemed to get no attention from Team Proxmox.

Today I tested a brand new Debian VM: Created in PVE 7, backed up to PBS, then restore on PVE again. That worked fine.

Restoring from PVE 6.4 backup is, unfortunately, still broken.

 

[YAGA]

Hello Fabian,

@fabian Please let me know if you would like, and if you have time, to check a non booting VM (bootloop in SeaBIOS) migrated from PVE 6.x to 7.x ?

Regards

 

[fabian]

how are the disks setup inside the VM? since seabios is seeing the disks, but a manual override doesn't work either, it must be that the actual disk is not bootable for seabios..

what happens if you insert an iso into the virtual CD rom drive, and boot a live cd and look at the disk? i.e., could you dump the partition tables of an old, broken VM and a new, working one?

 

[YAGA]

Hello Fabian,

Your test was very instructive.

I've added a live CD into the virtual CD rom drive on a non-booting VM. The live CD boots until the grub menu but it crashes immediately after selected the right setup in the grub menu...

After that, I've created a new VM with the live CD and this new VM boots on the live CD.

So, I've double-checked the differences with the non-booting VM and I found that the graphic card for the display was set to "serial terminal 0" for the non-booting VM.

This "serial terminal 0" setting was fine with PVE 6.x and old VM created with PVE6.x and reinstalled on PVE 7.0 until the beginning of September.

A PVE 7.x update in September or before didn't support the old created VM with the setting "serial terminal 0". It might be a bug or regression.

But now, if I recreate a new VM on PVE 7.0 with "serial terminal 0" setting it works.

To solve this issue, I've changed for the non-booting VM the "serial terminal 0" setting to defaults and it works.

Thank you very much for pointing out this idea to test with a live CD, you have made my day ;-)

Regards,

 

[vmware]

Hi all,

I can confirm that the SeaBIOS + serial terminal 0 issue has affected my migrated VMs, too. I believe that is why the issue isn't very widespread - Most people would've stick to the default, I suppose.

What will the bugfix look like? Will it fix the VMs that are currently "boot loop"? Thank you!

 

[YAGA]

how are the disks setup inside the VM? since seabios is seeing the disks, but a manual override doesn't work either, it must be that the actual disk is not bootable for seabios..

what happens if you insert an iso into the virtual CD rom drive, and boot a live cd and look at the disk? i.e., could you dump the partition tables of an old, broken VM and a new, working one?

Hi all,

I can confirm that the SeaBIOS + serial terminal 0 issue has affected my migrated VMs, too. I believe that is why the issue isn't very widespread - Most people would've stick to the default, I suppose.

What will the bugfix look like? Will it fix the VMs that are currently "boot loop"? Thank you!

It's a very good news for you too, I'm very happy for that.

Regards,

 

[vmware]

Further testing on PVE 7.0:

1. Created a new Debian 11 VM, worked OK with Display = Default
2. Shutdown VM; Added Serial Port (serial0); Changed display to "Serial terminal 0"
3. Start VM - "boot loop"

So the regression has nothing to do with backups, but VM display issue.

 

[YAGA]

how are the disks setup inside the VM? since seabios is seeing the disks, but a manual override doesn't work either, it must be that the actual disk is not bootable for seabios..

what happens if you insert an iso into the virtual CD rom drive, and boot a live cd and look at the disk? i.e., could you dump the partition tables of an old, broken VM and a new, working one?

@fabian Thanks again for your suggestion. Everything works fine now.

 

[lxiosjao]

Hi everyone,

Any idea why I have to select manually my disk OS in the console every reboot ?
If i dont do that, bootloop is happen.
In the boot option (GUI proxmox) there is only the possibility to put the pcie controller first (but not the disk).

SeaBios
Machine UUID
Select boot device :

1. AHCI/0 : Sandisk SSD
2. AHCI/1 : WDC ATA-9 hard-Disk
3. Legacy option rom
4. iPXE (PCI)