[SOLVED] PVE 8.0 Kernel crashes on Hetzner server

[bkraegelin]

Hi everyone,

I just ordered a new AX41 server at Hetzner and tried to install PVE 8.0.
Base is Debian 12 Bookworm, I use the installation from https://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_12_Bookworm. At least I try to...

pve-kernel-6.2 hangs on boot. I gave pve-kernel-6.1 a try, the same.

I analysed kernel boot messages and compared it with the messages from Debian kernel 6.1.0-9-amd64 which is the default after installation. From pve-kernel I see the following message (last message):

Jun 27 10:23:56 pve kernel: divide error: 0000 [#1] PREEMPT SMP NOPTI
Jun 27 10:23:56 pve kernel: CPU: 11 PID: 414 Comm: (udev-worker) Not tainted 6.2.16-3-pve #1
Jun 27 10:23:56 pve kernel: Hardware name: Hetzner /B450 Pro4 R2.0, BIOS H8.02C 05/15/2023
Jun 27 10:23:56 pve kernel: RIP: 0010:nvGetClocks+0x1bd/0x300 [nvidiafb]
Jun 27 10:23:56 pve kernel: Code: 0f 84 a6 00 00 00 3d 30 03 00 00 0f 84 9b 00 00 00 8b 8a 04 05 00 00 0f b6 c5 31 d2 41 0f af c1 44 0f b6 c9 c1 e9 10 83 e1 0f <41> f7 f1 d3 e8 41 89 00 48 8b 87 88 11 00 00 8b 88 00 05 00 00 0f

Followed by a stack trace.

Do you have any hints, what's going on?
Are there workarounds like disabling same hardware/drivers?
Might this issue be triggered by a BIOS bug?

Does anybody run PVE 8.0 successfully on Hetzner AX41?

Thanks for your help.
Birger

 

[bkraegelin]

Update:
Seems to be a bug in Nvidia driver [nvidiafb].

How can I disable that?

 

[bkraegelin]

Question to kernel gurus

Is it safe to disable the driver following this post?
https://linuxconfig.org/how-to-disa...idia-driver-on-ubuntu-20-04-focal-fossa-linux

Any help will be appreciated.
Thanks, Birger

 

[bkraegelin]

The kernel includes a buggy kernel module nvidiafb.

You may remove it using a blocklist and regenerating initramfs yourself.
Or - Proxmox Team removes it in the official kernel.

I did it, my Hetzner server boots flawlessly.

 

[herzkerl]

Did not work for me using the old Kernel. My AX41 still won't boot after installing Proxmox kernel.

 

[bkraegelin]

Did you block the nvidia driver as I indicated in my post? It's necessary.
The other way is to ignore the Proxmox Kernel and go with Debian Kernel.

My config as documented in my own wiki:
echo blacklist nouveau > /etc/modprobe.d/blacklist-nvidia-nouveau.conf echo options nouveau modeset=0 >> /etc/modprobe.d/blacklist-nvidia-nouveau.conf echo blacklist nvidiafb > /etc/modprobe.d/blacklist-nvidiafb.conf # update initramfs (use correct kernel version) update-initramfs -u -k 6.2.16-3-pve

 

[herzkerl]

First, I followed the Proxmox Wiki instructions:

echo "deb [arch=amd64] http://download.proxmox.com/debian/pve bullseye pve-no-subscription" > /etc/apt/sources.list.d/pve-install-repo.list
wget https://enterprise.proxmox.com/debian/proxmox-release-bullseye.gpg -O /etc/apt/trusted.gpg.d/proxmox-release-bullseye.gpg
apt update && apt full-upgrade
apt install pve-kernel-5.15

Then, I followed the instructions from the blog post:

sudo bash -c "echo blacklist nouveau > /etc/modprobe.d/blacklist-nvidia-nouveau.conf"
sudo bash -c "echo options nouveau modeset=0 >> /etc/modprobe.d/blacklist-nvidia-nouveau.conf"

cat /etc/modprobe.d/blacklist-nvidia-nouveau.conf
blacklist nouveau
options nouveau modeset=0

sudo update-initramfs -u

(Before installing pve-kernel-5.15 I did a reboot, which worked fine.)

 

[DNF-SaS]

Hi,

I have the same problem installing PVE 8.0-2 on a Hetzner AX41.
Any chance to install without first installing debian?

 

[herzkerl]

Hi,

I have the same problem installing PVE 8.0-2 on a Hetzner AX41.
Any chance to install without first installing debian?

I installed via installimage: https://forum.proxmox.com/threads/i...etzner-without-kvm-tutorial.94464/post-591024