🐋 Container lacks libnvidia-encode-* package

[uvayankee]

Is there an existing issue for this?

• I have searched the existing issues

Does this issue exist in the latest version?

• I'm using the latest release

Describe the bug?

When I stared the arm container with the proper Nvidia venc flags, it was still not listed as available, and transcodes failed. I was able to exec into the container and run HandbrakeCLI to see it complain of not being able to find libnvidia-encode.so.1 I was able to resolve the issue personally by running apt install libnivida-encode-

To Reproduce

1. pull the latest arm image on a machine with an nvidia card running the correct drivers
2. add --gpus all and -e NVIDIA_DRIVER_CAPABILITIES=all to your docker startup command
3. Start arm container
4. Note that Nvenc still has a red X

Built from source files ?

• I Built the image from source

What architecture are you seeing the problem on?

amd64

Expected behavior?

No need to install extra packages in the container to make it work

Relevant log output

No response

Anything else?

running apt install libnvidia-encode-470 (my driver version) solved this issue.

Code of Conduct

• I agree to follow this project's Code of Conduct

[github-actions]

If youre having issues, please remember to read the wiki and follow the instructions carefully

[microtechno9000]

Can you confirm your card supports NVENC?

https://github.com/automatic-ripping-machine/automatic-ripping-machine/wiki/nvidia

[shitwolfymakes]

I'll add this to the installs used in the arm-dependencies image

[shitwolfymakes]

This presents a problem, as the drivers differ based on user hardware (GTX 1XXX, RTX 2/3/4/5XXX, etc).

Can you:

• Delete and run the latest image
• Install nv-codec-headers in the container
• Try and transcode with NVENC

Please attach the logs when you do this. Additionally, what graphics card are you using?

[tim-nordell]

The docker image works here without doing any apt install commands.

...
    -e "NVIDIA_DRIVER_CAPABILITIES=all" \
    -e "NVIDIA_VISIBLE_DEVICES=all" \
    --gpus all \
    --runtime=nvidia \
...

Make sure you have the nVidia container toolkit installed, too, on your system. See https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installing-the-nvidia-container-toolkit.

[shitwolfymakes]

The docker image works here without doing any apt install commands.

...
    -e "NVIDIA_DRIVER_CAPABILITIES=all" \
    -e "NVIDIA_VISIBLE_DEVICES=all" \
    --gpus all \
    --runtime=nvidia \
...

Make sure you have the nVidia container toolkit installed, too, on your system. See https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installing-the-nvidia-container-toolkit.

Copy. In that case, could you PR this info against our documentation? I'd like to make sure you get the credit for figuring this out.

[uvayankee]

I will have to give that a try after I finish running my current batch.

My personal solution got complicated by some versioning issues along the way.

After upgrading Handbrake didn't accept my version of nvenc (too old) but going back to my old version caused MakeMKV to fail for being out of date.

I ended up writing my own Docker file to rationalize these issues, but it's certainly an edge case that won't be needed for almost anyone else.

I will close this after I test the new params above.

[tim-nordell]

@uvayankee -

I had the same issue about too old of a version of libencoder and had to update my version of the nvidia drivers on my main system (which was admittedly out of date). There's probably some minimum version for the container contents for the system side.