Is the GPG key used for signing RPMs available for download anywhere from a trusted source? The yum repo points to https://nvidia.github.io/nvidia-docker/gpgkey but that's not the key used to sign the current rpms.
Same problem for DEB repos as well.
Same here. Was working fine till about a day ago, but now get:
GPG error: https://nvidia.github.io/libnvidia-container/ubuntu18.04/amd64 InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY 6ED91CA3AC1160CD
E: The repository 'https://nvidia.github.io/libnvidia-container/ubuntu18.04/amd64 InRelease' is not signed.
I'm trying to use this in an offline environment where I need to manually create and maintain an offline repo. Generally it's not been an issue as I can use reposync to mirror the repo locally and construct a .repo file that points to a local copy of the gpg key for validation. The current public repo files for the Nvidia container packages disable gpg checks but that's not an option for the secure environment I need to deploy this in. Not to mention the uneasiness of something like docker that interacts at the kernel level running with unvalidated packages.
If I can download the key file locally then I can set up the mirrored repo to correctly validate offline. I just need the correct public gpg key to do it.
we (pytorch) are now getting widespread CI errors along the lines of:
E: Failed to fetch https://nvidia.github.io/nvidia-docker/ubuntu16.04/amd64/InRelease Clearsigned file isn't valid, got 'NOSPLIT' (does the network require authentication?)
The timing corresponds with these two commits:
https://github.com/NVIDIA/nvidia-docker/commit/6eed18c4c93215fbf076107c76b1ef7f48bb4e1f
https://github.com/NVIDIA/nvidia-docker/commit/295c69dacafce741f5e0995cabb273ef2ef9d9e8
@RenaudWasTaken are we doing something wrong here?
Hello!
@suo
The gpgkey expired, and we updated it with a new one :)
let me see if this isn't a caching issue
According to https://nvidia.github.io/nvidia-docker/ , gpgkey update can be done by:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | \
sudo apt-key add -
There are many issues that are aggregated here :)
Now that we've updated the repos (and the cache has been updated), you should just be able to add the new gpg key to your keyring and update / install the packages:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
@chuckatkins regarding the gpgkey not being the one used for signing, I didn't have time to check whether we used a different key (maybe the master), but with the key rotation the packages are now signed by the gpgkey advertised :)
Hope this solve everyone's issues!
Can I suggest then that the provided yum .repo files reenable validation then with gpgcheck=1?
current public repo files for the Nvidia container packages disable gpg checks but that's not an option for the secure environment I need to deploy this in.
That's not completly true, we enable repo_gpgcheck which allows you to ensure that the packages weren't tampered with.
I'm sifting through old documentation but I seem to remember that enabling gpgcheck was either breaking customers or us when generating the packages. I'll update here if I find more information.
According to https://nvidia.github.io/nvidia-docker/ , gpgkey update can be done by:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | \ sudo apt-key add -
That worked! Thanks.
So is this an expected process that will happen semi-often in future or totally randomly for some security reason?
According to https://nvidia.github.io/nvidia-docker/ , gpgkey update can be done by:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | \ sudo apt-key add -
This worked..thanks
Most helpful comment
According to https://nvidia.github.io/nvidia-docker/ , gpgkey update can be done by: