NVIDIA / nvidia-docker

Build and run Docker containers leveraging NVIDIA GPUs
Apache License 2.0
17.26k stars 2.03k forks source link

"public key is not available" for cuDNN APT repo in ubuntu 14.04 images #215

Closed grisaitis closed 8 years ago

grisaitis commented 8 years ago

I get the following error if I run apt-get update in one of the ubuntu 14.04-based images that includes cudnn:

W: GPG error: http://developer.download.nvidia.com  Release: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY F60F4B3D7FA2AF80

My Dockerfile looks like this:

FROM nvidia/cuda:7.5-cudnn5-devel-ubuntu14.04
RUN apt-get update

and I can get the same result if I replace nvidia/cuda:7.5-cudnn5-devel-ubuntu14.04 with cudnn4.

This error does not occur when I build from one of the ubuntu 16.04 Dockerfiles, e.g., nvidia/cuda:8.0-cudnn5-devel-ubuntu16.04. I believe the reason it works there is that cudnn isn't being installed from APT, but rather some tarball that's downloaded from nvidia.

grisaitis commented 8 years ago

I'm able to fix the issue by adding this before I run apt-get update:

RUN apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/7fa2af80.pub

I'll submit a pull request shortly with this change implemented.

grisaitis commented 8 years ago

Fixed the issue!

The error was caused by stale base images on my local docker host.

Re-pulling the images solved the problem, either with docker build --pull ... or for i in $(docker images --format "{{.Repository}}:{{.Tag}}" | grep nvidia/cuda); do docker pull $i; done.

grisaitis commented 8 years ago

To clarify: this was user error. No pull request / code changes warranted.

flx42 commented 8 years ago

No problem, sorry for breaking the public CUDA repo, the transition has been rough for everyone!

jacknlliu commented 7 years ago

I got the same error with Ubuntu 16.04. How to get the public key for ubuntu 16.04? @flx42

flx42 commented 7 years ago

You already asked on GitLab, and this issue has been fixed a long time ago, so you're facing a diffent issue.