search

NVIDIA / ansible-role-nvidia-driver

BSD 3-Clause "New" or "Revised" License
112 stars 64 forks source link

Problem to install the NVIDIA drivers on CentOS 8 #54

Closed AlanSSKim closed 2 years ago

AlanSSKim commented 2 years ago

Hi,

I've got a problem to install the NVIDIA Driver on CentOS8.

When installing for the first time, the latest version 495.29.05 was installed because the version was not specified. and then I have specified the driver version in default/main.yml as shown below.

nvidia_driver_package_state: present nvidia_driver_package_version: '470.57.02-1' nvidia_driver_persistence_mode_on: yes

after that I ran the playbook, but the installation failed as follows:

TASK [nvidia.nvidia_driver : install driver packages RHEL/CentOS 8 and newer] ***** fatal: [gdp-glm-gpu001]: FAILED! => changed=false msg: No group nvidia-driver:470.57.02-1 available. results: []

Even if I try changing it to multiple versions as shown below, it fails the same.

450.142.00 450.156.00-1 460.91.03-1 460.106.00-1 470.42.01-1 470.57.02-1

ajdecon commented 2 years ago

Due to changes in how the repositories are structured in RHEL/CentOS 8, the NVIDIA driver is now being installed as a YUM module rather than as a list of specific packages. This also changes the versioning scheme.

PR #53 adds support for specifying the module version more explicitly, and updates the docs in order to make it a bit clearer how to specify versions in RHEL 8 and similar distros.

I expect this PR to get reviewed and merged early next year, as most of the relevant team members are off for the US holiday break.