I recently attempted to configure one of my VMs on TrueNAS Scale 24.10 to use my NVIDIA Quadro 620 GPU, and something went haywire in the process, leading all of my docker instances managed by Dockge unable to start with the error Error response from daemon: unknown or invalid runtime name: nvidia (including ones that don’t reference the GPU). I’ve since disabled the GPU on that VM and rebooted, but I still can’t get any of my docker instances to start.
Also notable is that since this started, my system fails to run nvidia-smi from the command-line, erroring out with
‘$ nvidia-smi
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.’
It seems perhaps the driver needs to be reinstalled, but there’s no option in Apps > Installed > Configuration to uninstall/install/reinstall the driver. Is there a way to force a reinstall from command line?
I could be wrong, but going to Apps> Configuration> Settings & then unchecking the “install nvidia drivers’ & then ‘save’ should remove them. Check it again & save to re-install.
Otherwise, you can always back up your config & re-install; it is pretty much painless.
I just did two things:
- Updated to 25.10 (finally), after unchecking “Install NVIDIA drivers” under
Apps > Configuration > Settings which I assume would uninstall the drivers?
- Replaced my NVIDIA P620 with an Intel A310 Eco
I now have no references to the NVIDIA runtime in any of my docker compose files (managed with Dockge) and some of them will start, while others (with no mention of nvidia in the config) will still fail to start with
Error response from daemon: unknown or invalid runtime name: nvidia
Is the only way around this to do a clean install from scratch, or is there something else I can do to clear this up?
Install & importing config isn’t a ‘start from scratch’; I’ve had to do it multiple times to do various issues caused by my own curiousity & I’ve yet to notice anything missing. It honestly may be faster than ttying to figure out a different solution… just don’t forget to backup config first