@ghthor, thanks for your solution recommendation. Maybe @kiendang would consider it for the new version?
@kiendang, congratulations for being maintainer for this package:)
Git Clone URL: | https://aur.archlinux.org/nvidia-container-runtime.git (read-only, click to copy) |
---|---|
Package Base: | nvidia-container-runtime |
Description: | NVIDIA opencontainer runtime fork to expose GPU devices to containers. |
Upstream URL: | https://github.com/NVIDIA/nvidia-container-toolkit |
Keywords: | docker nvidia runc |
Licenses: | Apache |
Submitter: | lukeyeager |
Maintainer: | kiendang |
Last Packager: | kiendang |
Votes: | 28 |
Popularity: | 0.32 |
First Submitted: | 2017-09-12 21:38 (UTC) |
Last Updated: | 2023-07-19 14:36 (UTC) |
@ghthor, thanks for your solution recommendation. Maybe @kiendang would consider it for the new version?
@kiendang, congratulations for being maintainer for this package:)
@orkut @mrksr @Palmik
See my comment. I've managed to get nvidia-container-{toolkit,runtime} working by removing -buildmode=pie
from each of the packages builds.
https://aur.archlinux.org/packages/nvidia-container-toolkit#comment-942940
@Palmik, have you solved your problem with PKGBUILD?
edit: I've solved mine with PKGBUILD. Thank you @mrksr for the link:)
@Palmik there is some discussion on gitlab. It includes a working PKGBUILD that can be used until this package is updated.
Did anyone see this error? When running nvidia-container-runtime
:
nvidia-container-runtime: error while loading shared libraries: unexpected PLT reloc type 0x00
I also tried testing with docker directly using docker run --rm --runtime=nvidia --gpus all nvidia/cuda:11.6.2-base-ubuntu20.04 nvidia-smi
, which results in:
docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: unable to retrieve OCI runtime error (open /run/containerd/io.containerd.runtime.v2.task/moby/d21f793fce6d10eb027f402e6f6f672c33f8230019f1d7c528dfbbf767d47d28/log.json: no such file or directory): nvidia-container-runtime did not terminate successfully: exit status 127: unknown.
These are the relevant package versions:
local/cuda 12.2.0-1
local/cudnn 8.9.2.26-1
local/egl-wayland 2:1.1.12-1
local/lib32-nvidia-utils 535.104.05-1
local/libnvidia-container 1.13.5-1
local/libnvidia-container-tools 1.13.5-1
local/libvdpau 1.5-2
local/nvidia 535.104.05-2
local/nvidia-container-runtime 3.13.1-4
local/nvidia-container-toolkit 1.13.5-1
local/nvidia-docker 2.13.0-1
local/nvidia-utils 535.104.05-1
local/nvtop 3.0.2-1
local/opencl-nvidia 535.104.05-1
base-devel
package is assumed to be installed when using the AUR.
line 48: patch: command not found
after adding sudo pacman -S --needed base-devel
it built
The linking issue should be fixed now. Apologies for the delay, it was a weird thing to track down and my typical computer I use to work on this stuff is in storage at the moment.
apologies for the delay on fixing this, I'm currently packing to move across the US but will have time later this week to try and look at it.
Any progress with the linking issue?
Pinned Comments
jshap commented on 2019-03-20 15:16 (UTC)
This package was changed to properly build from source. If you want to continue using prebuilt binaries you can use
nvidia-container-runtime-bin
.