Hello,
I’m trying to follow the instructions in this almalinux blog post It was just posted about 1 month ago so I would expect it to be correct. but the second dnf command errors out with the following:
# dnf install nvidia-open-kmod nvidia-driver
AlmaLinux 9 - CRB 41 kB/s | 4.2 kB 00:00
Error:
Problem: package nvidia-driver-3:580.82.07-1.el9.x86_64 from cuda-rhel9-x86_64 requires nvidia-kmod-common = 3:580.82.07, but none of the providers can be installed
- package nvidia-kmod-common-3:580.82.07-1.el9.noarch from cuda-rhel9-x86_64 requires nvidia-kmod = 3:580.82.07, but none of the providers can be installed
- cannot install the best candidate for the job
- package kmod-nvidia-latest-dkms-3:580.82.07-1.el9.x86_64 from cuda-rhel9-x86_64 is filtered out by exclude filtering
- package kmod-nvidia-open-dkms-3:580.82.07-1.el9.noarch from cuda-rhel9-x86_64 is filtered out by exclude filtering
(try to add '--skip-broken' to skip uninstallable packages or '--nobest' to use not only best candidate packages)
The machine I’m installing on is a headless server with a new install of Almalinux 9.6. I hope to also install the container toolkit to allow docker containers to share the GPU.
Any suggestions would be appreciated. I have not seen any other posts about this.
I believe the “exclude” entry in “nvidia-cuda.repo” is causing the issue.
Try removing it or commenting it out. The installation should succeed.
[root@alma9 yum.repos.d]# dnf install nvidia-open-kmod nvidia-driver
Last metadata expiration check: 0:02:35 ago on Thu Sep 4 17:49:09 2025.
Error:
Problem: package nvidia-driver-3:580.82.07-1.el9.x86_64 from cuda-rhel9-x86_64 requires nvidia-kmod-common = 3:580.82.07, but none of the providers can be installed
- package nvidia-kmod-common-3:580.82.07-1.el9.noarch from cuda-rhel9-x86_64 requires nvidia-kmod = 3:580.82.07, but none of the providers can be installed
- cannot install the best candidate for the job
- package kmod-nvidia-latest-dkms-3:580.82.07-1.el9.x86_64 from cuda-rhel9-x86_64 is filtered out by exclude filtering
- package kmod-nvidia-open-dkms-3:580.82.07-1.el9.noarch from cuda-rhel9-x86_64 is filtered out by exclude filtering
(try to add '--skip-broken' to skip uninstallable packages or '--nobest' to use not only best candidate packages)
[root@alma9 yum.repos.d]# cat nvidia-cuda.repo
# This repository is for NVIDIA user space and CUDA components which cannot be redistributed due to licensing restrictions.
# This content is served directly by NVIDIA.
# Do not modify this file, it will be overwritten/replaced on updates.
[cuda-rhel$releasever-$basearch]
name=cuda-rhel$releasever-$basearch
baseurl=https://developer.download.nvidia.com/compute/cuda/repos/rhel$releasever/$basearch
enabled=1
gpgcheck=1
countme=1
gpgkey=https://developer.download.nvidia.com/compute/cuda/repos/rhel$releasever/$basearch/D42D0685.pub
exclude=akmod-nvidia dkms-nvidia kmod-nvidia-*-dkms nvidia-fs* nvidia-gds* cuda-drivers
[root@alma9 yum.repos.d]# pwd
/etc/yum.repos.d
[root@alma9 yum.repos.d]# cat nvidia-cuda.repo
# This repository is for NVIDIA user space and CUDA components which cannot be redistributed due to licensing restrictions.
# This content is served directly by NVIDIA.
# Do not modify this file, it will be overwritten/replaced on updates.
[cuda-rhel$releasever-$basearch]
name=cuda-rhel$releasever-$basearch
baseurl=https://developer.download.nvidia.com/compute/cuda/repos/rhel$releasever/$basearch
enabled=1
gpgcheck=1
countme=1
gpgkey=https://developer.download.nvidia.com/compute/cuda/repos/rhel$releasever/$basearch/D42D0685.pub
exclude=akmod-nvidia dkms-nvidia kmod-nvidia-*-dkms nvidia-fs* nvidia-gds* cuda-drivers
I should have mentioned that I tried that already. It errors also:
# vim /etc/yum.repos.d/nvidia-cuda.repo
[root@homehost1 ~]# dnf install nvidia-open-kmod nvidia-driver
cuda-rhel9-x86_64 34 kB/s | 3.5 kB 00:00
Error:
Problem: cannot install both nvidia-kmod-common-3:580.82.07-1.el9.noarch from cuda-rhel9-x86_64 and nvidia-kmod-common-3:580.65.06-1.el9.noarch from cuda-rhel9-x86_64
- package nvidia-driver-3:580.82.07-1.el9.x86_64 from cuda-rhel9-x86_64 requires nvidia-kmod-common = 3:580.82.07, but none of the providers can be installed
- package nvidia-open-kmod-580.65.06-2.el9.x86_64 from almalinux-nvidia requires nvidia-kmod-common = 3:580.65.06, but none of the providers can be installed
- cannot install the best candidate for the job
- conflicting requests
(try to add '--allowerasing' to command line to replace conflicting packages or '--skip-broken' to skip uninstallable packages or '--nobest' to use not only best candidate packages)
A thing is that the NVidia’s repo ( cuda-rhel9-x86_64) has all packages for one driver version. That is, including the kernel modules. The additional AlmaLinux repo has only package with kernel modules.
Why kernel module from Alma, when NVidia offers it too? Because Alma’s module is signed by Alma and thus requires no extra steps when Secure Boot is enabled.
Since we need kernel module only once, we don’t want to install the other copy. The excludes are there to help in that.
Do you have Secure Boot enabled?
mokutil --sb
should tell.
The DNF does cache metadata about repos and occasionally that gets stale. One can clear/flush things:
dnf clean all
What do you have and see now?
dnf list \*nvidia\*
Will this be just for computation (with CUDA applications), no GUI desktop sessions?
If yes, then you don’t need all the drivers, just the ones that CUDA needs.
(It is possible to install multiple versions of CUDA simultaneously, if applications are prebuilt to depend on specific version. One does not need whole toolkit for those, cuda-runtime-X-Y ought to suffice.)
does anyone know how to use an old card requiring 470xx drivers (k5200) with 9.7? i had it working with 9.6, upgrade broke it… i guess the alternative is nouveau?
Which driver were you using? (As in “CUDA repo”, “ELRepo”, “RPM Fusion” …)
ELRepo has kmod-nvidia-470xx in their testing repo (aka elrepo-testing) and
there is already version 470.256.02-7.1.el9_7.elrepo
Even the versions for el9_6 are only in the testing.