Running Proxmox on an old HP Proliant server, newest kernel fails to boot

[samip537]

Hi there,

I started off my Proxmox install with installing Debian Buster and then proxmox on top of that as guided at: https://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_Buster.

However, the Debian kernel works fine but pve-kernel does not. Somewhere between version vmlinuz-5.4.30-1-pve and vmlinuz-5.4.106-1-pve has been some changes which renders it not bootable. It will only show a black screen without much information as to what's wrong.

For reference, Debian kernel that does boot is 4.19.0-14-amd64.

Hardware is an HP ProLiant DL140 G3, which has two Intel Xeon 5130 CPUs.

 

[ness1602]

Maybe try with 5.11?

 

[samip537]

Maybe try with 5.11?

Just did that, and that also does not boot.

 

[Ramalama]

2 Options:

1.
Example only: GRUB_CMDLINE_LINUX_DEFAULT="quiet console=tty0 console=ttyS0,115200n8"
/etc/default/grub
- add "console=tty0 console=ttyS0,115200n8" to your CMDLINE.
- remove quiet, but SOL shows everything anyway.
- activate SOL in ILO and connect over SOL, i don't know if your super old Java ilo can record, thats why install ipmitools simply and connect to your ILO and record the SOL output (to a textfile). Or maybe easier, you can ssh into your ilo with putty, there is an option to enter sol in the cli too.
You need to google a bit, i don't remember which commands it was to get the output in the ssh cli of ilo.
- Post the recorded textfile (bootlog) here, then we can see what is missing and where it hangs.

2.
If it's too complicated with SOL, probably a screenshot where is stucks is helpfull too (if you get the java remote console working xD)
- but this isn't that super helpfull, cause some of the bootlogging is hidden sometimes.
- you need to remove quiet from cmdline, cause else we won't see where it hangs.

Cheers

 

[0xd149e38e]

DL140 G3

I wonder how old it is. 15 years?

Edit:
The thing is: Stuff gets removed from the kernel as well. This happend to my old scsci controller, since "nobody was was using it anymore".

 

[samip537]

2 Options:

1.
Example only: GRUB_CMDLINE_LINUX_DEFAULT="quiet console=tty0 console=ttyS0,115200n8"
/etc/default/grub
- add "console=tty0 console=ttyS0,115200n8" to your CMDLINE.
- remove quiet, but SOL shows everything anyway.
- activate SOL in ILO and connect over SOL, i don't know if your super old Java ilo can record, thats why install ipmitools simply and connect to your ILO and record the SOL output (to a textfile). Or maybe easier, you can ssh into your ilo with putty, there is an option to enter sol in the cli too.
You need to google a bit, i don't remember which commands it was to get the output in the ssh cli of ilo.
- Post the recorded textfile (bootlog) here, then we can see what is missing and where it hangs.

2.
If it's too complicated with SOL, probably a screenshot where is stucks is helpfull too (if you get the java remote console working xD)
- but this isn't that super helpfull, cause some of the bootlogging is hidden sometimes.
- you need to remove quiet from cmdline, cause else we won't see where it hangs.

Cheers

My iLO does not have such functionality to access serial and it also does not have SSH but the system has on-board serial connector which I don't have a cable for.

I wonder how old it is. 15 years?

Edit:
The thing is: Stuff gets removed from the kernel as well. This happend to my old scsci controller, since "nobody was was using it anymore".

BIOS version is from 2005-2006.

 

[Ramalama]

Well, without logs, no one can help you, sorry.

Find a way, you have still the option to remove "quiet" and make a screenshot or picture.

 

[0xd149e38e]

You can try if an old version of Proxmox works for you. See: http://download.proxmox.com/iso/

 

[samip537]

Okey, I managed to find a working kernel, but it's 5.0.8-2.

I tried the following: 5.4.106, 5.4.30 , 5.0.21.5 and 5.0.8-2.
Only one that booted was the 5.0.8-2.

 

[0xd149e38e]

That's pretty good! You've got a running Proxmox.

I'd suggest you try it out first, to see if it works good enough for you on this server. A lot has changed in terms of virtualization support in last 15 years. Also your server supports only up to 16GB RAM, this might be limiting.

If you are happy with it, then you can investigate further: You know the problem was introduced somewhere between kernels 5.0.8-2 and 5.0.21-5.