Hello guys!
I’m trying setup an headless RL server for computation workloads.
The server is equipped with an NVIDIA GPU and it runs RL 8.4 without graphical desktop environment.
I’m trying to install the NVIDIA GPU dirvers to enable CUDA support without installing the graphical desktop environment too.
How can I do it?
Thanks for help.
Ivan
NVidia has Yum-repositories that provide CUDA toolkit and drivers properly RPM-packaged: Installation Guide Linux :: CUDA Toolkit Documentation
PS. There is no “RL 8.4”. There is “RL 8” and it has currently content based on RHEL 8.5.
If you still have content from 8.4 release, then do sudo dnf up ASAP.
Thanks for response.
As you can see, following the procedure suggested by NVIDIA (already done many times) there’s a lot of graphical dependencies (e.g., xorg server) to be installed.
# dnf module install nvidia-driver:latest-dkms
Last metadata expiration check: 0:00:51 ago on Sat 19 Feb 2022 09:38:40 AM EST.
Dependencies resolved.
=============================================================================================================================================================
Package Architecture Version Repository Size
=============================================================================================================================================================
Installing group/module packages:
cuda-drivers x86_64 510.47.03-1 cuda-rhel8-x86_64 7.0 k
kmod-nvidia-latest-dkms x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 28 M
nvidia-driver x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 22 M
nvidia-driver-NVML x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 516 k
nvidia-driver-NvFBCOpenGL x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 52 k
nvidia-driver-cuda x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 591 k
nvidia-driver-cuda-libs x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 63 M
nvidia-driver-devel x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 12 k
nvidia-driver-libs x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 168 M
nvidia-kmod-common noarch 3:510.47.03-1.el8 cuda-rhel8-x86_64 12 k
nvidia-libXNVCtrl x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 25 k
nvidia-libXNVCtrl-devel x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 55 k
nvidia-modprobe x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 36 k
nvidia-persistenced x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 42 k
nvidia-settings x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 832 k
nvidia-xconfig x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 105 k
Installing dependencies:
adwaita-cursor-theme noarch 3.28.0-2.el8 appstream 646 k
adwaita-icon-theme noarch 3.28.0-2.el8 appstream 11 M
at-spi2-atk x86_64 2.26.2-1.el8 appstream 88 k
at-spi2-core x86_64 2.28.0-1.el8 appstream 168 k
colord-libs x86_64 1.4.2-1.el8 appstream 234 k
dkms noarch 3.0.3-1.el8 epel 62 k
dnf-plugin-nvidia noarch 2.0-1.el8 cuda-rhel8-x86_64 12 k
egl-wayland x86_64 1.1.7-1.el8 appstream 33 k
gtk3 x86_64 3.22.30-8.el8 appstream 4.5 M
libX11-devel x86_64 1.6.8-5.el8 appstream 975 k
libX11-xcb x86_64 1.6.8-5.el8 appstream 13 k
libXau-devel x86_64 1.0.9-3.el8 appstream 19 k
libXdmcp x86_64 1.1.3-1.el8 appstream 40 k
libXfont2 x86_64 2.0.3-2.el8 appstream 147 k
libXtst x86_64 1.2.3-7.el8 appstream 21 k
libdrm x86_64 2.4.106-2.el8 appstream 166 k
libepoxy x86_64 1.5.8-1.el8 appstream 224 k
libevdev x86_64 1.10.0-1.el8 appstream 43 k
libglvnd x86_64 1:1.3.2-1.el8 appstream 126 k
libglvnd-egl x86_64 1:1.3.2-1.el8 appstream 48 k
libglvnd-gles x86_64 1:1.3.2-1.el8 appstream 39 k
libglvnd-glx x86_64 1:1.3.2-1.el8 appstream 136 k
libglvnd-opengl x86_64 1:1.3.2-1.el8 appstream 46 k
libinput x86_64 1.16.3-2.el8 appstream 216 k
libpciaccess x86_64 0.14-1.el8 baseos 31 k
libvdpau x86_64 1.4-2.el8 appstream 40 k
libwacom x86_64 1.6-3.el8 appstream 41 k
libwacom-data noarch 1.6-3.el8 appstream 103 k
libwayland-client x86_64 1.19.0-1.el8 appstream 38 k
libwayland-cursor x86_64 1.19.0-1.el8 appstream 25 k
libwayland-egl x86_64 1.19.0-1.el8 appstream 18 k
libwayland-server x86_64 1.19.0-1.el8 appstream 46 k
libxcb-devel x86_64 1.13.1-1.el8 appstream 1.1 M
libxkbfile x86_64 1.1.0-1.el8 appstream 87 k
libxshmfence x86_64 1.3-2.el8 appstream 12 k
llvm-libs x86_64 12.0.1-2.module+el8.5.0+692+8756646f appstream 23 M
mesa-libEGL x86_64 21.1.5-1.el8.0 appstream 134 k
mesa-libGL x86_64 21.1.5-1.el8.0 appstream 183 k
mesa-libgbm x86_64 21.1.5-1.el8.0 appstream 56 k
mesa-libglapi x86_64 21.1.5-1.el8.0 appstream 64 k
mesa-vulkan-drivers x86_64 21.1.5-1.el8.0 appstream 6.1 M
mtdev x86_64 1.1.5-12.el8 appstream 23 k
ocl-icd x86_64 2.2.12-1.el8 appstream 50 k
opencl-filesystem noarch 1.0-6.el8 appstream 7.3 k
rest x86_64 0.8.1-2.el8 appstream 69 k
vulkan-loader x86_64 1.2.198.0-2.el8_5 appstream 122 k
xorg-x11-drv-fbdev x86_64 0.5.0-2.el8 appstream 26 k
xorg-x11-drv-libinput x86_64 0.29.0-1.el8 appstream 49 k
xorg-x11-drv-vesa x86_64 2.4.0-3.el8 appstream 30 k
xorg-x11-proto-devel noarch 2020.1-3.el8 appstream 279 k
xorg-x11-server-Xorg x86_64 1.20.11-2.el8 appstream 1.5 M
xorg-x11-server-common x86_64 1.20.11-2.el8 appstream 41 k
xorg-x11-xkb-utils x86_64 7.7-28.el8 appstream 113 k
Installing weak dependencies:
dconf x86_64 0.28.0-4.el8 appstream 107 k
Installing module profiles:
nvidia-driver/default
Enabling module streams:
llvm-toolset rhel8
nvidia-driver latest-dkms
Transaction Summary
=============================================================================================================================================================
Install 70 Packages
Total download size: 336 M
Installed size: 836 M
Is this ok [y/N]:
I’m searching for a procedure for installing CUDA without graphical stuffs.
Thank you.
Ivan
If you look at the in the installation guide jlehtone linked there are instructions in the advanced-setup section of the guide.
sudo dnf install cuda-toolkit-11-6 nvidia-driver-cuda akmod-nvidia
Hi @tkuraku!
I’ve already tried:
# dnf install cuda-toolkit-11-6 nvidia-driver-cuda akmod-nvidia
Last metadata expiration check: 0:36:40 ago on Sat 19 Feb 2022 02:39:27 PM EST.
No match for argument: akmod-nvidia
Error: Unable to find a match: akmod-nvidia
Next I found out the package kmod-nvidia-latest-dkms:
# dnf info kmod-nvidia-latest-dkms
Last metadata expiration check: 0:02:14 ago on Sat 19 Feb 2022 03:24:05 PM EST.
Available Packages
Name : kmod-nvidia-latest-dkms
Epoch : 3
Version : 510.47.03
Release : 1.el8
Architecture : x86_64
Size : 28 M
Source : kmod-nvidia-latest-dkms-510.47.03-1.el8.src.rpm
Repository : cuda-rhel8-x86_64
Summary : NVIDIA display driver kernel module
URL : http://www.nvidia.com/object/unix.html
License : NVIDIA License
Description : This package provides the proprietary Nvidia kernel driver
: modules. The modules are rebuilt through the DKMS system when a
: new kernel or modules become available.
I think this could replace akmod-nvidia.
The problem is that when try to install packages, it seems Xorg will be installed, but I don’t need it on an headless server.
# dnf install cuda-toolkit-11-6 nvidia-driver-cuda kmod-nvidia-latest-dkms
Last metadata expiration check: 0:08:42 ago on Sat 19 Feb 2022 03:24:05 PM EST.
Dependencies resolved.
=============================================================================================================================================================
Package Architecture Version Repository Size
=============================================================================================================================================================
Installing:
cuda-toolkit-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 4.2 k
kmod-nvidia-latest-dkms x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 28 M
nvidia-driver-cuda x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 591 k
Installing dependencies:
alsa-lib x86_64 1.2.5-4.el8 appstream 488 k
cuda-cccl-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 1.7 M
cuda-command-line-tools-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 2.9 k
cuda-compiler-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 2.7 k
cuda-cudart-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 215 k
cuda-cudart-devel-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 1.3 M
cuda-cuobjdump-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 134 k
cuda-cupti-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 23 M
cuda-cuxxfilt-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 276 k
cuda-documentation-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 121 k
cuda-driver-devel-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 35 k
cuda-gdb-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 5.2 M
cuda-libraries-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 2.9 k
cuda-libraries-devel-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 3.1 k
cuda-license-10-1 x86_64 10.1.243-1 cuda-rhel8-x86_64 22 k
cuda-memcheck-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 172 k
cuda-nsight-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 113 M
cuda-nsight-compute-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 5.2 k
cuda-nsight-systems-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 4.9 k
cuda-nvcc-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 47 M
cuda-nvdisasm-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 31 M
cuda-nvml-dev-10-1 x86_64 10.1.243-1 cuda-rhel8-x86_64 70 k
cuda-nvml-devel-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 99 k
cuda-nvprof-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 4.5 M
cuda-nvprune-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 67 k
cuda-nvrtc-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 20 M
cuda-nvrtc-devel-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 100 M
cuda-nvtx-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 83 k
cuda-nvvp-11-6 x86_64 11.6.58-1 cuda-rhel8-x86_64 115 M
cuda-samples-11-6 x86_64 11.6.101-1 cuda-rhel8-x86_64 33 k
cuda-sanitizer-11-6 x86_64 11.6.55-1 cuda-rhel8-x86_64 17 M
cuda-toolkit-11-6-config-common noarch 11.6.55-1 cuda-rhel8-x86_64 3.1 k
cuda-toolkit-11-config-common noarch 11.6.55-1 cuda-rhel8-x86_64 3.4 k
cuda-toolkit-config-common noarch 11.6.55-1 cuda-rhel8-x86_64 3.4 k
cuda-tools-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 2.7 k
cuda-visual-tools-11-6 x86_64 11.6.0-1 cuda-rhel8-x86_64 2.9 k
dkms noarch 3.0.3-1.el8 epel 62 k
dnf-plugin-nvidia noarch 2.0-1.el8 cuda-rhel8-x86_64 12 k
egl-wayland x86_64 1.1.7-1.el8 appstream 33 k
gds-tools-11-6 x86_64 1.2.0.100-1 cuda-rhel8-x86_64 40 M
giflib x86_64 5.1.4-3.el8 appstream 50 k
java-1.8.0-openjdk-headless-slowdebug x86_64 1:1.8.0.322.b06-2.el8_5 powertools 36 M
java-1.8.0-openjdk-slowdebug x86_64 1:1.8.0.322.b06-2.el8_5 powertools 345 k
libX11-xcb x86_64 1.6.8-5.el8 appstream 13 k
libXdmcp x86_64 1.1.3-1.el8 appstream 40 k
libXfont2 x86_64 2.0.3-2.el8 appstream 147 k
libXtst x86_64 1.2.3-7.el8 appstream 21 k
libcublas-11-6 x86_64 11.8.1.74-1 cuda-rhel8-x86_64 314 M
libcublas-devel-11-6 x86_64 11.8.1.74-1 cuda-rhel8-x86_64 326 M
libcufft-11-6 x86_64 10.7.0.55-1 cuda-rhel8-x86_64 98 M
libcufft-devel-11-6 x86_64 10.7.0.55-1 cuda-rhel8-x86_64 206 M
libcufile-11-6 x86_64 1.2.0.100-1 cuda-rhel8-x86_64 627 k
libcufile-devel-11-6 x86_64 1.2.0.100-1 cuda-rhel8-x86_64 13 M
libcurand-11-6 x86_64 10.2.9.55-1 cuda-rhel8-x86_64 51 M
libcurand-devel-11-6 x86_64 10.2.9.55-1 cuda-rhel8-x86_64 52 M
libcusolver-11-6 x86_64 11.3.2.55-1 cuda-rhel8-x86_64 110 M
libcusolver-devel-11-6 x86_64 11.3.2.55-1 cuda-rhel8-x86_64 70 M
libcusparse-11-6 x86_64 11.7.1.55-1 cuda-rhel8-x86_64 158 M
libcusparse-devel-11-6 x86_64 11.7.1.55-1 cuda-rhel8-x86_64 322 M
libdrm x86_64 2.4.106-2.el8 appstream 166 k
libepoxy x86_64 1.5.8-1.el8 appstream 224 k
libevdev x86_64 1.10.0-1.el8 appstream 43 k
libglvnd x86_64 1:1.3.2-1.el8 appstream 126 k
libglvnd-egl x86_64 1:1.3.2-1.el8 appstream 48 k
libglvnd-gles x86_64 1:1.3.2-1.el8 appstream 39 k
libglvnd-glx x86_64 1:1.3.2-1.el8 appstream 136 k
libglvnd-opengl x86_64 1:1.3.2-1.el8 appstream 46 k
libinput x86_64 1.16.3-2.el8 appstream 216 k
libnpp-11-6 x86_64 11.6.0.55-1 cuda-rhel8-x86_64 122 M
libnpp-devel-11-6 x86_64 11.6.0.55-1 cuda-rhel8-x86_64 119 M
libnvjpeg-11-6 x86_64 11.6.0.55-1 cuda-rhel8-x86_64 2.6 M
libnvjpeg-devel-11-6 x86_64 11.6.0.55-1 cuda-rhel8-x86_64 2.3 M
libpciaccess x86_64 0.14-1.el8 baseos 31 k
libvdpau x86_64 1.4-2.el8 appstream 40 k
libwacom x86_64 1.6-3.el8 appstream 41 k
libwacom-data noarch 1.6-3.el8 appstream 103 k
libwayland-client x86_64 1.19.0-1.el8 appstream 38 k
libwayland-server x86_64 1.19.0-1.el8 appstream 46 k
libxkbfile x86_64 1.1.0-1.el8 appstream 87 k
libxshmfence x86_64 1.3-2.el8 appstream 12 k
llvm-libs x86_64 12.0.1-2.module+el8.5.0+692+8756646f appstream 23 M
mesa-libEGL x86_64 21.1.5-1.el8.0 appstream 134 k
mesa-libGL x86_64 21.1.5-1.el8.0 appstream 183 k
mesa-libgbm x86_64 21.1.5-1.el8.0 appstream 56 k
mesa-libglapi x86_64 21.1.5-1.el8.0 appstream 64 k
mesa-vulkan-drivers x86_64 21.1.5-1.el8.0 appstream 6.1 M
mtdev x86_64 1.1.5-12.el8 appstream 23 k
nsight-compute-2022.1.0 x86_64 2022.1.0.12-1 cuda-rhel8-x86_64 458 M
nsight-systems-2021.5.2 x86_64 2021.5.2.53_28d0e6e-0 cuda-rhel8-x86_64 226 M
nvidia-driver x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 22 M
nvidia-driver-cuda-libs x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 63 M
nvidia-driver-libs x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 168 M
nvidia-kmod-common noarch 3:510.47.03-1.el8 cuda-rhel8-x86_64 12 k
nvidia-persistenced x86_64 3:510.47.03-1.el8 cuda-rhel8-x86_64 42 k
ocl-icd x86_64 2.2.12-1.el8 appstream 50 k
opencl-filesystem noarch 1.0-6.el8 appstream 7.3 k
ttmkfdir x86_64 3.0.9-54.el8 appstream 61 k
vulkan-loader x86_64 1.2.198.0-2.el8_5 appstream 122 k
xorg-x11-drv-fbdev x86_64 0.5.0-2.el8 appstream 26 k
xorg-x11-drv-libinput x86_64 0.29.0-1.el8 appstream 49 k
xorg-x11-drv-vesa x86_64 2.4.0-3.el8 appstream 30 k
xorg-x11-fonts-Type1 noarch 7.5-19.el8 appstream 521 k
xorg-x11-server-Xorg x86_64 1.20.11-2.el8 appstream 1.5 M
xorg-x11-server-common x86_64 1.20.11-2.el8 appstream 41 k
xorg-x11-xkb-utils x86_64 7.7-28.el8 appstream 113 k
Enabling module streams:
llvm-toolset rhel8
nvidia-driver latest-dkms
Transaction Summary
=============================================================================================================================================================
Install 108 Packages
Total download size: 3.4 G
Installed size: 6.7 G
Is this ok [y/N]:
At this point I don’t know if it is possible to install CUDA and NVIDIA driver without install Xorg.
Thanks for help.
Ivan
Some of its packages, but not all of Xorg. Yes, they will consume some disk space – not much (it is the CUDA and NVidia driver that devour disk) – and your setup remains headless.
Ok, message got.
I installed using:
# dnf install cuda-toolkit-11-6 nvidia-driver-cuda kmod-nvidia-latest-dkms
blacklisted nouveau, regenerated initramfs, rebooted.
Now, when I run nvidia-smi I obtain the following message:
# nvidia-smi
NVIDIA-SMI couldn't find libnvidia-ml.so library in your system. Please make sure that the NVIDIA Display Driver is properly installed and present in your system.
Please also try adding directory that contains libnvidia-ml.so to your system PATH.
What’s going on?
Ivan
I’ve not yet solved, any ideas?