[SOLVED] Can't reboot any VM

Charles Osman

Member
Apr 4, 2017
8
0
6
Hi,

I've installed a cluster of 5 servers for testing purpose using Proxmox 5 Beta updated a few minutes ago. Both Centos and debian based VMs I created fail to reboot. I must first stop the VM before being able to start it again.
reboot and shutdown -r now commands from the VM fail.
Reset and CTRL+ALT+DEL from the vnc fail.
The VM restart but hang with a prompt at the top left corner of the screen. Right after selecting the any entry from grub.

Is this something that can cause this issue: https : // review.coreboot.org/cgit/seabios.git/commit/?id=c68aff57ce317d9f2d69d20eba893a10d964f316

Thank you,
Charlie
 
Hi,

Same issue using UEFI :(
I can reboot the VMs by disabling ACPI option. But in this case I can't add multiple cores to the VMs.

Any advise will be welcome,
Thank you.
Charlie
 
Encouraged to hear I'm not the only one bitten by this. I get the same symptoms on a Dell poweredge r730 with BIOS Version: 2.0.1.

Charles - disabling ACPI on the kernel line of grub, or in the proxmox VM options don't seem to help for me. In fact disabling ACPI through proxmox on a ubuntu test VM now kernel panics reliably (regardless of it's kernel boot line), and my Debian test VM still fails to reboot..

I'd love to hear a way forward on this as VM's doing a single task then rebooting is a stable of my setup!
A colleague had suggested a BIOS upgrade, but wanted to check what hw you were running before doing that route.

Side note. I'm considering buying a subscription, but want to make sure this is fixed first. Could I install updated packages from a future Beta image / other source, or would I need a subscription in order to unlock the pve enterprise repo for fixes here?
 
Hi David,

I am testing on Dell PowerEdge R630 updated with the latest firmware and Bios. Configuring one socket and one core works for me too.

Thank you,
Charlie
 
Testing the SeaBIOS theory:

I tried compiling a new version of SeaBIOS from git clone git://git.seabios.org/seabios.git seabios
Dropped it into place at /usr/share/kvm/bios-256k.bin
And it didn't fix the behaviour - although dmidecode on the guests did show it using the new BIOS 1.10 something and NOT the 1.9.3 prebuilt qemu project one bundled.

So either: The BIOS patch theory is the wrong culprit, or the patch isn't effective (my checked out code had the patch). I see it refers to Xen/Qemu, so not sure if that also encompasses kvm (my C is rubbish).
 
For giggles I tried the seabios-256 version from my working 4.something proxmox install (didn't work), and even the entire /usr/share/kvm directory from there in case of dependencies, and it still didn't behave, similar behaviour but more hanging on the proxmox splash boot screen..
 
qemu packages based on 2.9-rc3 should hit the PVE 5.0 pvetest repositories some time this week and should fix the reboot issue
 
  • Like
Reactions: David Harvey
Thanks for the update Fabian. If you have any test packages I'm happy to dpkg -i them on and try it out. I was moments away from paving my 5beta install and rolling back to 4.4, so if there are any other workarounds I could try to avoid nuking my install that would be most useful!