Debian buster: nnvidia-container-cli: ldcache error: process /sbin/ldconfig failed with error code: 1

[tobycheese]

1. Issue or feature description

docker run --runtime=nvidia --rm nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04 nvidia-smi

results in

docker: Error response from daemon: OCI runtime create failed: container_linux.go:348: starting container process caused "process_linux.go:402: container init caused \"process_linux.go:385: running prestart hook 1 caused \\\"error running hook: exit status 1, stdout: , stderr: exec command: [/usr/bin/nvidia-container-cli --load-kmods --debug=/var/log/nvidia-container-runtime-hook.log configure --ldconfig=@/sbin/ldconfig --device=all --compute --utility --require=cuda>=9.0 --pid=2004 /var/lib/docker/overlay2/0765a24c941dbb710987be1a3ac397bc09cbfeac26c2d1f8f91f1434f91e0af2/merged]\\\\nnvidia-container-cli: ldcache error: process /sbin/ldconfig failed with error code: 1\\\\n\\\"\"": unknown.

2. Steps to reproduce the issue

I installed on Debian buster, but adding the repo for stretch. Is this bound to fail? nvidia-docker1 was working just fine. Do I need to downgrade, or is buster support here soon?

3. Information to attach (optional if deemed irrelevant)

• Kernel version: Linux epsilon 4.14.0-3-amd64 Add README image #1 SMP Debian 4.14.17-1 (2018-02-14) x86_64 GNU/Linux
• Driver information from nvidia-smi -a: 390.25
• Docker version from docker version: 17.12.1-ce
• NVIDIA packages version from dpkg -l '*nvidia*' or rpm -qa '*nvidia*'
• NVIDIA container library version from nvidia-container-cli -V
• NVIDIA container library logs (see troubleshooting)

nvidia-packages.txt
nvidia-container-library-version.txt
nvidia-container-runtime-hook.log

[flx42]

I have a fix for this, will test

[flx42]

Issue was fixed in NVIDIA/libnvidia-container@cef6c8f, but it's not released as a package yet.
You will need to make && make install the project from sources, or better make deb and dpkg -i dist/libnvidia-container1[...]

[tobycheese]

Thanks for the quick fix! I tried building from source but it fails due to not finding cuda.h. I have not installed cuda on my host, that's what I am using containers for. I couldn't quickly figure out how to install cuda on debian, the .run files from nvidia don't list ubuntu, and apt-get install nvidia-cuda-toolkit wants to install 256 packages, which I am not going to do on my host. I also tried building inside a docker container, but either I can't install cuda or can't get gcc-7 (what the host uses).

Guess I'll downgrade to nvidia-docker1 and wait for an updated package, which is something I can live with :-) Thanks again!

[flx42]

Ah, sorry. Try "make docker-debian", if I'm not mistaken.

…

On Sat, Mar 10, 2018, 15:04 tobycheese ***@***.***> wrote: Thanks for the quick fix! I tried building from source but it fails due to not finding cuda.h. I have not installed cuda on my host, that's what I am using containers for. I couldn't quickly figure out how to install cuda on debian, the .run files from nvidia don't list ubuntu, and apt-get install nvidia-cuda-toolkit wants to install 256 packages, which I am not going to do on my host. I also tried building inside a docker container, but either I can't install cuda or can't get gcc-7 (what the host uses). Guess I'll downgrade to nvidia-docker1 and wait for an updated package, which is something I can live with :-) Thanks again! — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#660 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADegjcbACfB_DoxK6vQp75kvB43yv3z-ks5tdFvlgaJpZM4Sk-fV> .

[OrBin]

@flx42 When is the fix expected to be released?

[Trophime]

Hi, trying to rebuild from scratch with make docker-ubuntu18.04 fails:

dpkg-deb: error: parsing file 'debian/nvidia-docker2/DEBIAN/control' near line 6 package 'nvidia-docker2':
 'Depends' field, reference to 'docker-ce': error in version: version number does not start with digit
dh_builddeb: dpkg-deb --build debian/nvidia-docker2 .. returned exit code 2
dh_builddeb: Aborting due to earlier error
debian/rules:7: recipe for target 'binary' failed
make: *** [binary] Error 2

same for debian flavor...