Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[voidlinux-github]

[-- Attachment #1: Type: text/plain, Size: 1196 bytes --]

New comment by gmorenz on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-570110482

Comment:
I tried installing these packages, they installed successfully but running `sudo docker run --gpus all nvidia/cuda:9.0-base nvidia-smi` (as suggested by [nvidia's documentation](https://github.com/NVIDIA/nvidia-docker#usage)) failed with the following error message.

```
$ sudo docker run --gpus all nvidia/cuda:9.0-base nvidia-smi
docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused "process_linux.go:449: container init caused \"process_linux.go:432: running prestart hook 0 caused \\\"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: detection error: open failed: /usr/lib/libnvoptix.so.1: no such file or directory\\\\n\\\"\"": unknown.
ERRO[0000] error waiting for container: context canceled 
```

I fixed this by running `cd /usr/lib; sudo ln -s libnvoptix.so.440.44 libnvoptix.so.1`, after which the command worked successfully.

Symlinking x.so.440.44 to x.so.1 feels wrong... so I'm not sure if this is the *right* fix... but it does work.

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[voidlinux-github]

[-- Attachment #1: Type: text/plain, Size: 1091 bytes --]

New comment by shaeri on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-570148966

Comment:
Creating the symlink is correct and usually libfoo.so and libfoo.so.1 are symlinks to the libfoo..so.version. ldconfig usually takes care of creating the symlinks. That is part of what ldconfig does. So after installing the nvidia drivers (nvidia440.44) you should run ldconfig as sudo user and it will take care of creating all the symlinks.

`sudo ldconfig`

The issue here is that libnvoptix is part of the nvidia driviers and somehow the symlinks to libnvoptix.so and libnvoptix.so.1 are not created as part of driver installation. The nvidia drivers are not part of the free void-linux repos and furthermore installing them without creating the correct xorg.conf may break the display, therfore, I opted for not including them to be installed alongside these 2 packages.

I can include the last step of running ldconfig as a post installation step. However, if nvidia drivers are installed after these 2 packages, this problem will persist.

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[voidlinux-github]

[-- Attachment #1: Type: text/plain, Size: 1127 bytes --]

New comment by shaeri on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-570148966

Comment:
Thanks for testing the packages.

Creating the symlink is correct and usually libfoo.so and libfoo.so.1 are symlinks to the libfoo..so.version. ldconfig usually takes care of creating the symlinks. That is part of what ldconfig does. So after installing the nvidia drivers (nvidia440.44) you should run ldconfig as sudo user and it will take care of creating all the symlinks.

`sudo ldconfig`

The issue here is that libnvoptix is part of the nvidia driviers and somehow the symlinks to libnvoptix.so and libnvoptix.so.1 are not created as part of driver installation. The nvidia drivers are not part of the free void-linux repos and furthermore installing them without creating the correct xorg.conf may break the display, therfore, I opted for not including them to be installed alongside these 2 packages.

I can include the last step of running ldconfig as a post installation step. However, if nvidia drivers are installed after these 2 packages, this problem will persist.

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[voidlinux-github]

[-- Attachment #1: Type: text/plain, Size: 1124 bytes --]

New comment by shaeri on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-570148966

Comment:
Thanks for testing the packages.

Creating the symlink is correct and usually libfoo.so and libfoo.so.1 are symlinks to the libfoo..so.version. ldconfig usually takes care of creating the symlinks. That is part of what ldconfig does. So after installing the nvidia drivers (nvidia440.44) you should run ldconfig as sudo user and it will take care of creating all the symlinks.

`sudo ldconfig`

The issue here is that libnvoptix is part of the nvidia driviers and somehow the symlinks libnvoptix.so and libnvoptix.so.1 are not created as part of driver installation. The nvidia drivers are not part of the free void-linux repos and furthermore installing them without creating the correct xorg.conf may break the display, therfore, I opted for not including them to be installed alongside these 2 packages.

I can include the last step of running ldconfig as a post installation step. However, if nvidia drivers are installed after these 2 packages, this problem will persist.

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[voidlinux-github]

[-- Attachment #1: Type: text/plain, Size: 1165 bytes --]

New comment by shaeri on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-570148966

Comment:
Thanks for testing the packages.

Creating the symlink is correct and usually libfoo.so and libfoo.so.1 are symlinks to the libfoo..so.version. ldconfig usually takes care of creating the symlinks. That is part of what ldconfig does. So after installing the nvidia drivers (nvidia440.44) you should run ldconfig as sudo user and it will take care of creating all the symlinks.

`sudo ldconfig`

The issue here is that libnvoptix is part of the nvidia driviers and somehow the symlinks libnvoptix.so and libnvoptix.so.1 are not created as part of driver installation. The nvidia drivers are not part of the free void-linux repos and furthermore installing them without creating the correct xorg.conf may break the display, therfore, I opted for not including them to be installed alongside these 2 packages.

I can include the execution of ldconfig as a post installation step in the libnvidia-container installation template. However, if nvidia drivers are installed after these 2 packages, this problem will persist.

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[vixus0]

[-- Attachment #1: Type: text/plain, Size: 178 bytes --]

New comment by vixus0 on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-744049727

Comment:
Any idea when this might get reviewed?

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[subtlevenom]

[-- Attachment #1: Type: text/plain, Size: 395 bytes --]

New comment by subtlevenom on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-958577996

Comment:
works well, no errors. Tested with:
sudo docker run --gpus all nvidia/cuda:9.0-base nvidia-smi
cker run --gpus all -it --rm tensorflow/tensorflow:latest-gpu python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"

Re: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[github-actions]

[-- Attachment #1: Type: text/plain, Size: 305 bytes --]

New comment by github-actions[bot] on void-packages repository

https://github.com/void-linux/void-packages/pull/17505#issuecomment-1099786930

Comment:
Pull Requests become stale 90 days after last activity and are closed 14 days after that.  If this pull request is still relevant bump it or assign it.

Re: [PR PATCH] [Closed]: New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…

[github-actions]

[-- Attachment #1: Type: text/plain, Size: 1133 bytes --]

There's a closed pull request on the void-packages repository

New packages: libnvidia-container-1.0.5 and nvidia-container-runtime-…
https://github.com/void-linux/void-packages/pull/17505

Description:
…3.1.4

There are 2 packages added in order to enable using Nvidia devices in
containers. This has been tested with docker containers.
Two packages are included in one commit since the nvidia-container-runtime
package depends on libnvidia-container.

This commit resolves: #11084.
Even though the package requst is for the nvidia-docker
(https://github.com/NVIDIA/nvidia-docker) package,
as of Docker 19.03, docker recognize GPU devices, therefore, the nvidia-docker
library is not required anymore (https://github.com/NVIDIA/nvidia-docker).
Since the current version of docker on the void-packages is 19.03.5,
the nvidia-docker library would not be required to be included in the void packages.

libnvidia-container (https://github.com/NVIDIA/libnvidia-container) and
nvidia-container-runtime (https://github.com/NVIDIA/nvidia-container-runtime)
are required to be installed to add nvidia runtime to Docker.