Please use the thread below to discuss the related blog post:
Hello,
Have you tried the update HostOS flow and is there some guide on steps rebuild the drivers with the new kernel modules?
Reason I ask is that we are trying to integrate this solution into our product, and actions like driver updates work ok, but the moment that we update the Host OS, even after the drivers get rebuilt, is impossible to insert them.
Regards,
Is there any more convenient way to get Nvidia drivers and stuff running?
Seems that in 5.x kernels unloading Nouveau module causes an exception (infamous nvkm_falcon_v1_wait_for_halt), leaving GPU in some intermediate state and crashing Nvidia original drivers.
For a user other than root inside the container, the GPU access did not work. I had to add the user to the nvidia-persistenced group.
RUN usermod -a -G nvidia-persistenced <user>