System information:
Operating System: RHEL
CPE OS Name: cpe:/o:redhat:enterprise_linux:7.9:GA:server
Kernel: Linux 3.10.0-1160.11.1.el7.x86_64
Architecture: x86-64
GPU: 4x NVIDIA Titan X (Pascal)
Hi, I’ve spoken with Red Hat about my questions below, but they didn’t have specific answers, and they suggest that I contacted NVIDIA support.
My server has RHEL 7.9 set up on it. I would like to figure out if it is possible to install nvidia-docker on it so that I can run containerized-GPU compute jobs on it.
I have several questions:
-
In reviewing Red Hat’s instructions for installing CUDA drivers for RHEL 7, they indicate that the NVIDIA driver is also required. Is this driver required if you are not displaying onto a monitor (e.g., connecting to the RHEL server for remote GPU compute)?
NOTE: Nvidia-docker specifies that the NVIDIA driver is required, whereas the CUDA driver is not required. -
The NVIDIA-docker toolkit currently specifies that it only supports up to RHEL 7.7. Do you know if it is possible to get NVIDIA-docker working on RHEL 7.9?
-
Do you know if the Docker version support on RHEL 7.9 (as shown by these instructions getting Docker on RHEL 7) includes Docker >= 19.03? This version range is indicated as a recommended requirement for nvidia-docker.
-
Do you have any other recommendations or suggested instructions for getting nvidia-docker set up on a RHEL server for remote GPU compute? Is a different version of RHEL 7 needed to make this possible?
Note, I initially submitted an email query and was told by Lohith in Customer Care that I needed to post my question here.
Thank you.
Resources:
- Red Hat, installing CUDA on RHEL 7: https://access.redhat.com/solutions/1453633
- nvidia-docker: https://github.com/NVIDIA/nvidia-docker
- NVIDIA docker toolkit shows RHEL support up to RHEL 7.7: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#linux-distributions