`nvidia-docker` is deprecated?

We are helping a client use our License Plate Recognition (ALPR) software (https://platerecognizer.com). We have been long supporters of Nvidia devices with our ALPR software. Our ALPR software runs on a Docker container.

However, today, when one of our clients tried to install our ALPR, we received the following error message when installing docker on their JetPack 4.4.

jt: ~ ] nvidia-docker run --runtime nvidia --rm -t -p 8080:8080 -v license:/license -e TOKEN=xxxxxxxxxxxxxxxxxx -e LICENSE_KEY=xxxxxxxxxx platerecognizer/alpr-jetson
docker: Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused “process_linux.go:449: container init caused “process_linux.go:432: running prestart hook 1 caused \“error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: driver error: failed to process request\\n\”””: unknown.

From our client: “Nvidia says nvidia-docker is deprecated. It looks like they relatively recently amped-up their Docker support and things changed. We’ve been unable to figure out the legacy dependencies. We are using the latest Nano disk images pulled this week directly from Nvidia, so we’re as vanilla as can be.”

As context,

  • docker run hello-world --> runs as expected.
  • dpkg-query --show nvidia-l4t-core --> nvidia-l4t-core 32.4.2-20200408182156
  • docker run --runtime nvidia --> it returns this error: “docker: Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused “process_linux.go:449: container init caused “process_linux.go:432: running prestart hook 1 caused \“error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: driver error: failed to process request\\n\”””: unknown.”

Any help or advice to resolve this?

You can use the --gpus flag with docker run instead. See usage below.
https://github.com/NVIDIA/nvidia-docker#usage

Edit: i see you’re referring to Tegra. On Tegra you can use --runtime nvidia Not sure what is causing your error when trying that. Are you on TK1 or Nano?

Client is on Jetson Nano.

@KyleT Please see this thread for a possible solution:

re: deprecation, I think that only applies to x86 (for now).

@TomK Can you move this thread to the Nano forum please?

1 Like

Hi,

Thanks for reporting this to us.

It’s looks like the OS version from your client is incorrect.
It should be 32.4.3 if they are using JetPack4.4 GA released recently.

$ dpkg-query --show nvidia-l4t-core
nvidia-l4t-core	32.4.3-20200625213407

We can also run the docker run --runtime nvidia without issue on the rel-32.4.3 base.

$ docker run --runtime nvidia
"docker run" requires at least 1 argument.
See 'docker run --help'.

Usage:  docker run [OPTIONS] IMAGE [COMMAND] [ARG...]

Run a command in a new container

Thanks.