Intagrate the deployment of the Intel gpu plugin into to the dss snap

[misohu]

Why it needs to get done

We don't want users to manually install Intel gpu plugin with kustomize commands . It would be better from UX to be able to install Intel gpu plugin directly with one command from DSS.

What needs to get done

The following are things we need to explore for the technical design of the functionality:

• Should the user specify the plugin version to install? How will the user know which version to install?
• Decide what to do if the plugin is already presented in the cluster (with same version - should abort ? , older version - should upgrade? , newer version should abort?.
  • Can we even upgrade the installed plugin?
  • How can we remove the installed plugin?
• Design a way for dss to install Intel gpu plugin with one command

When is the task considered done

• The spec for the solution is approved
• Dss command for installing Intel gpu plugin is implemented.

[syncronize-issues-to-jira]

Thank you for reporting us your feedback!

The internal ticket has been created: https://warthogs.atlassian.net/browse/KF-6181.

This message was autogenerated

[frenchwr]

For detecting what plugin version is installed you can use something like:

kubectl get pod intel-gpu-plugin-9bswt -o jsonpath='{.spec.containers[0].image}'
intel/intel-gpu-plugin:0.30.0

Note the suffix in the pod name varies between deployments but you should be able to determine it programmatically.

[frenchwr]

For upgrading that is something that we should test. I know the operator supports in-place upgrades of the various plugins it manages, which makes me think the same is true for the plugins (e.g. the GPU plugin) themselves.