Nvidia-driver-latest on Centos7 Fails to install yum-plugin-nvidia-0.5-1.el7.no

Pretty straight forward…

Run a docker container with centos:centos7

Configure the repos for epel and nvidia
yum install https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm

ARCH=x86_64
distribution=rhel7
yum-config-manager --add-repo http://developer.download.nvidia.com/compute/cuda/repos/distribution/{ARCH}/cuda-$distribution.repo

Then attempt to install the driver

yum install nvidia-driver-latest (or nvidia-driver-latest-dkms or cuda-10-2)

It will end up with

yum-plugin-nvidia-0.5-1.el7.no FAILED
http://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/yum-plugin-nvidia-0.5-1.el7.noarch.rpm: [Errno -1] Package does not match intended download. Suggestion: run yum --enablerepo=cuda clean metadata
Trying other mirror.

Running the yum cleanup, etc, doesn’t seem to help.

I’m not totally certain when this started happening, but I noticed it yesterday evening.

1 Like

I have the same issue and first encountered it last saturday.

The previous job did not have this issue, which means it manifested in the time interval

2020-06-06T02:34+02:00/2020-06-06T14:36+02:00

i install yum-plugin-nvidia-0.5 fine today on CentOS 7.7 (we doing large roll-out)

with
sudo yum install yum-plugin-nvidia-0.5

That seems to fail for me, wants to install 0.5-1.el7. I installed yum-plugin-nvidia-0.3-1.el7.noarch and yum update fails, along with yum install yum-plugin-nvidia-0.5

# yum install yum-plugin-nvidia-0.5
Loaded plugins: fastestmirror, nvidia, ovl
#### NVIDIA ####
Loading mirror speeds from cached hostfile

Resolving Dependencies
--> Running transaction check
---> Package yum-plugin-nvidia.noarch 0:0.3-1.el7 will be updated
---> Package yum-plugin-nvidia.noarch 0:0.5-1.el7 will be an update
--> Finished Dependency Resolution

Dependencies Resolved

===============================================================================================================================================================================================================================================================================
 Package                                                                   Arch                                                           Version                                                           Repository                                                    Size
===============================================================================================================================================================================================================================================================================
Updating:
 yum-plugin-nvidia                                                         noarch                                                         0.5-1.el7                                                         cuda                                                         9.8 k

Transaction Summary
===============================================================================================================================================================================================================================================================================
Upgrade  1 Package

Total download size: 9.8 k
Is this ok [y/d/N]:y
Downloading packages:
Delta RPMs disabled because /usr/bin/applydeltarpm not installed.
yum-plugin-nvidia-0.5-1.el7.no FAILED                                                                                     0% [                                                                                                               ]  0.0 B/s |    0 B  --:--:-- ETA
http://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/yum-plugin-nvidia-0.5-1.el7.noarch.rpm: [Errno -1] Package does not match intended download. Suggestion: run yum --enablerepo=cuda clean metadata                       ]  0.0 B/s |    0 B  --:--:-- ETA
Trying other mirror.


Error downloading packages:
  yum-plugin-nvidia-0.5-1.el7.noarch: [Errno 256] No more mirrors to try.

I just downloaded directly yum-plugin-nvidia-0.5-1.el7.noarch.rpm and installed it with ‘rpm -i’. Then I rerean ‘yum install cuda-drivers’ and it worked.

The error message suggests to clean the metadata, did you try that?

The error message suggests to clean the metadata, did you try that?

As I have the same issue with the same error message: I tried that, doesn’t help (not that I expected it to, since I already do a yum clean all beforehand).

I checked: The metadata in the repository isn’t right. When they updated the package, They updated the filename in the metadata, but not the the sha256sum. Since the sha256 sum of the downloaded file doesn’t match whats in the repository metadata, yum assumes you’ve downloaded a corrupt or malicious package and rejects it.

sha256sum yum-plugin-nvidia-0.5-1.el7.noarch.rpm

36da817af7bd286c3af60b0dbae3bc43d4c804ed4bda52ef4d2f10a202a98387 yum-plugin-nvidia-0.5-1.el7.noarch.rpm

ls -l yum-plugin-nvidia-0.5-1.el7.noarch.rpm

-rw-r–r--. 1 root root 10112 Jun 2 20:53 yum-plugin-nvidia-0.5-1.el7.noarch.rpm

from repo metadata
hash: 62fdbd314908d82852783f3fc638e574c78ce4cc1eb02f9c426f8fc91ddd2813
size: 10056

My install is completely stalled until they fix their repository.

1 Like

Well spotted. Maybe email it to linux-bugs[at]nvidia.com for additional attention.

I’ve sent off what I’ve discovered to that team, hopefully it’s resolved soon.

If there’s a checksum mismatch, that could indicate a failed supply chain attack too. I’d be careful installing something that doesn’t match signatures.

I thought the same, and it’s good advice so I haven’t taken any steps to force install until I hear back from Nvidia.

But I have done some more digging for data, and found that the sha256sum of the package in the repository does match older repo meta still available in /repodata. This suggests perhaps I got it backwards, maybe the repo meta got updated with a new sha256sum but the new package was never pushed.

I’ve hit this same error.

Nvidia has acknowledged my email and this seems like it will be quite easy to fix on their end.
Hopefully they’ll get it fixed up soon.

2 Likes

AFAICT the repo has been fixed.

1 Like

Was just about to say the same.

I have ran into some minor discrepancies with how the curand library is being treated with a libcurand-11-0-10.2.0.191-1.x86_64.rpm… not sure why cuda 11 stuff is getting installed when I’m only using 10-2 rpms.

I ended up adding a yum install cuda-10-2-curand and seems like things are okay. I assume they removed a dependency that was installing it for me or it just behaves differently.