nvidia-docker
nvidia-docker copied to clipboard
NVIDIA
Support Jetson Xavier NX
1. Issue or feature description
Are there any plans to support also the Jetson Xavier NX device?
Just want to bump this issue as I have a similar sentiment though for Jetson AGX Xavier.
Hey, this is working well for me on a Xavier NX, initially I had some issues and so I tried starting with the JetPack image again and all working fine since.
Hey, this is working well for me on a Xavier NX, initially I had some issues and so I tried starting with the JetPack image again and all working fine since.
@reubenbijl what kind of issues were you having? At first, I noticed that the nvidia-docker2 came pre-installed with the Jetpack 4.2 so I did the "hello-world" shell script and it went fine. But when I'm trying to create a container using nvidia/cuda, I'm getting the error below:
nvidia@x02:~$ nvidia-docker run --rm -ti nvidia/cuda nvcc --version
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.
For reference, I also posted my issue on this post where people have a similar problem but no solutions yet: https://github.com/NVIDIA/nvidia-container-toolkit/issues/183
Yes, very similar 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 0 caused \"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: driver error: failed to process request\\n\""": unknown."
I didn't note everything I did on my first install compared to the second, but one major difference was setting up via GNOME and configuring RDP, whereas the second configuration I used the serial interface to configure the device, and only used it via ssh.
Interesting.. I'm accessing it via SSH as well, yet I'm having issues. May I know what's the Jetpack version of your working environment?
cat /etc/nv_tegra_release
R32 (release), REVISION: 4.3, GCID: 21589087, BOARD: t186ref, EABI:
aarch64, DATE: Fri Jun 26 04:34:27 UTC 2020
On Sat, Aug 8, 2020 at 11:14 AM paldana-ISI [email protected] wrote:
Interesting.. I'm accessing it via SSH as well, yet I'm having issues. May I know what's the Jetpack version of your working environment?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NVIDIA/nvidia-docker/issues/1353#issuecomment-670779700, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOMN66NULI4FUG52JZAUQTR7SDGTANCNFSM4PEPI3AQ .
apt-cache show nvidia-jetpack
Package: nvidia-jetpack
Version: 4.4-b186
Architecture: arm64
On Sat, Aug 8, 2020 at 11:16 AM Reuben Bijl [email protected] wrote:
cat /etc/nv_tegra_release
R32 (release), REVISION: 4.3, GCID: 21589087, BOARD: t186ref, EABI:
aarch64, DATE: Fri Jun 26 04:34:27 UTC 2020
On Sat, Aug 8, 2020 at 11:14 AM paldana-ISI [email protected] wrote:
Interesting.. I'm accessing it via SSH as well, yet I'm having issues. May I know what's the Jetpack version of your working environment?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NVIDIA/nvidia-docker/issues/1353#issuecomment-670779700, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOMN66NULI4FUG52JZAUQTR7SDGTANCNFSM4PEPI3AQ .
Yes, same image.
On Sat, Aug 8, 2020 at 11:19 AM paldana-ISI [email protected] wrote:
Thanks! Was your first install also using Jetpack 4.4?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NVIDIA/nvidia-docker/issues/1353#issuecomment-670780866, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOMN6YPTIJNNFTOTJASNQ3R7SDZ3ANCNFSM4PEPI3AQ .
Hey, this is working well for me on a Xavier NX, initially I had some issues and so I tried starting with the JetPack image again and all working fine since.
@reubenbijl what kind of issues were you having? At first, I noticed that the nvidia-docker2 came pre-installed with the Jetpack 4.2 so I did the "hello-world" shell script and it went fine. But when I'm trying to create a container using nvidia/cuda, I'm getting the error below:
nvidia@x02:~$ nvidia-docker run --rm -ti nvidia/cuda nvcc --version
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.For reference, I also posted my issue on this post where people have a similar problem but no solutions yet: #1203 (comment)
I am having the same issue with my Jetson AGX Xavier 32Gb, @paldana-ISI any luck on following the second fresh install mentioned? I am wondering if I should go ahead and try it.
This particular error is often encountered when one is trying to run containers for a different architecture. Looking at the nvidia/cuda repo on docker hub confirms this - there are only amd64 images published. You may be able to build an arm64 version directly on the Xavier board reusing the same Dockerfile though.
FYI, there are ARM images, but they're not under a different tag, they have a different name:
sudo nvidia-docker run --net=host --rm nvidia/cuda-arm64:11.0-base nvidia-smi
Looks the Jetson Xavier AGX doesn't come with the nvidia docker stuff pre-installed. If you use SDK Manager to install that, and you run the ARM image, then it might work.
The problem has been solved. Share it with you：
sudo apt install libnvidia-container-tools libnvidia-container0:arm64 nvidia-container-runtime nvidia-container-toolkit nvidia-docker2
The newer versions of the NVIDIA Container Toolkit offer improved Tegra support.
If problems persist, please open an issue against https://github.com/NVIDIA/nvidia-container-toolkit
