Setting up nvidia drivers

[nawhouse]

Hello.
I have only been using proxmox for a few months now and I was working on setting up a host that has a nvidia Tesla P4 GPU in it. I want to get vGPU up and running but I keep running into issues installing the nvidia enterprise drivers. I have tried the most recent drivers as well as some slightly older ones and I keep running into the same problem. I am attaching the log files from the install. If anyone could help that would be super.

Thanks

 

[BobhWasatch]

There are build errors compiling the driver. Seems that some struct definitions are not as expected. This usually means you have the wrong version of the driver for the kernel you are building for.

It appears that Release 470 is going on three years old so it would almost certainly need an update for the 6.8 kernel PVE uses. I do not use NVidia myself so can't offer anything more specific.

 

[nawhouse]

There are build errors compiling the driver. Seems that some struct definitions are not as expected. This usually means you have the wrong version of the driver for the kernel you are building for.

It appears that Release 470 is going on three years old so it would almost certainly need an update for the 6.8 kernel PVE uses. I do not use NVidia myself so can't offer anything more specific.

Thanks.
I have tired drivers from the most current to the older 470 version and all get the same issues.

 

[leesteken]

I have tired drivers from the most current to the older 470 version and all get the same issues.

Only NVidia can provide updates to their (closed-source) drivers. Maybe ask them for support for Proxmox/Ubuntu linux kernel 6.8?

 

[FlemingC]

Only NVidia can provide updates to their (closed-source) drivers. Maybe ask them for support for Proxmox/Ubuntu linux kernel 6.8?

NVIDIA will not provide support for Proxmox, only for Ubuntu/Red Hat. Even if you fudge and say Debian, that is not supported either. This was back in February and I could be wrong and they may have modified that policy

 

[leesteken]

NVIDIA will not provide support for Proxmox, only for Ubuntu/Red Hat. Even if you fudge and say Debian, that is not supported either. This was back in February and I could be wrong and they may have modified that policy

Support for Ubuntu is good enough as Proxmox uses a Ubuntu kernel. Proxmox or the community cannot support NVidia since their driver is not open-source and they cannot make fixes to it. So I guess NVidia users are sore out of luck...

 

[FlemingC]

Support for Ubuntu is good enough as Proxmox uses a Ubuntu kernel. Proxmox or the community cannot support NVidia since their driver is not open-source and they cannot make fixes to it. So I guess NVidia users are sore out of luck...

Fortunately for me, the issue I had was more of a licensing configuration and NVIDIA support was able to help me there. But from the start they told me they do not provide "support" for Proxmox nor Debian-based installations, which Proxmox is, isn't it?

 

[leesteken]

Fortunately for me, the issue I had was more of a licensing configuration and NVIDIA support was able to help me there. But from the start they told me they do not provide "support" for Proxmox nor Debian-based installations, which Proxmox is, isn't it?

Yes, Proxmox is based on Debian with a Ubuntu (also based on Debian) kernel. I'm sorry if you get stuck between Proxmox (who can't fix NVidia drivers) and NVidia (who won't support the devices you bought). Personally, I think the way out is to buy hardware from a different vendor or use an alternative to Proxmox (which is supported by NVidia).

 

[FlemingC]

Yes, Proxmox is based on Debian with a Ubuntu (also based on Debian) kernel. I'm sorry if you get stuck between Proxmox (who can't fix NVidia drivers) and NVidia (who won't support the devices you bought). Personally, I think the way out is to buy hardware from a different vendor or use an alternative to Proxmox (which is supported by NVidia).

Yes, we can't afford the VMware tax so we are looking more closely at Ubuntu/Red Hat as the basis for an Openstack, CloudStack, or OpenNebula cloud

 

[leesteken]

Yes, we can't afford the VMware tax

I guess you need to make sure that the Proxmox linux kernel version you are running is supported by NVidia's driver. Maybe stay on kernel version 6.5 or a while (since lots of threads on this forum appear to show that a recent Nvidia driver works fine with it). Alternatively, switch to AMD or Intel GPUs which have native open-source support and come with each linux kernel version automaticelly. They are free, except that people need to pay driver developers to keep the opnr-source drivers freely available.