intel-device-plugins-for-kubernetes icon indicating copy to clipboard operation
intel-device-plugins-for-kubernetes copied to clipboard

Instructions or link for reviewing what plugin READMEs tell to install

Open eero-t opened this issue 2 years ago • 7 comments

In GPU plugin README install section: https://github.com/intel/intel-device-plugins-for-kubernetes/tree/main/cmd/gpu_plugin#deploy-with-pre-built-container-image

Both links to releases give 404:

  • https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/gpu_plugin?ref=v.0.24.0
  • https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/gpu_plugin/overlays/nfd_labeled_nodes?ref=v0.24.0

For some reason "kubectl apply -k" works e.g. for first link, but it's better if one can review what is to be installed, before actually doing it.

eero-t avatar Jun 22 '22 08:06 eero-t

kubectl apply -k --dry-run can be used for viewing what is going to be constructed as a result of kustomization of that directory, those links are not supposed to be viewed in browser.

byako avatar Jul 07 '22 11:07 byako

kubectl apply -k --dry-run can be used for viewing what is going to be constructed as a result of kustomization of that directory, those links are not supposed to be viewed in browser.

That tells just type and name of the created objects:

$ kubectl apply --dry-run=client -k https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/gpu_plugin?ref=v0.24.0
daemonset.apps/intel-gpu-plugin created (dry run)

And nothing about their properties: what they require, where they are deployed, what accesses are granted to the pods etc, like link here would tell: https://github.com/intel/intel-device-plugins-for-kubernetes/tree/release-0.24/deployments

PS. "desired release version" could be a link to: https://github.com/intel/intel-device-plugins-for-kubernetes/releases

eero-t avatar Jul 20 '22 14:07 eero-t

@eero-t byako wasn't perhaps descriptive enough above. With -o yaml in the end you can see the content.

Try kubectl apply --dry-run=client -k https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/gpu_plugin?ref=v0.24.0 -o yaml

In any case the links weren't intended for browser viewing

uniemimu avatar Jul 20 '22 14:07 uniemimu

Ok, that works. Maybe something like this could be added to end of the section (along with above mentioned link to releases page):

PS. To review what above kubectl commands would apply to your cluster, try them first with --dry-run=client -o yaml options.

?

eero-t avatar Jul 20 '22 14:07 eero-t

What the other plugins do:

  • Example for applying only from Git checkout: IIA, VPU, FPGA
  • Examples both for applying from URL and Git checkout: SGX (using operator), QAT, GPU, DLB

@mythi this is somewhat inconsistent. It would be good for all plugins to have similar instructions (and documentation structure). For URL installations, above --dry-run=client -o yaml note would be nice.

eero-t avatar Jul 20 '22 14:07 eero-t

I'm always using

kubectl kustomize https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/gpu_plugin?ref=v0.24.0

mythi avatar Jul 20 '22 17:07 mythi

@mythi this is somewhat inconsistent.

yep, the docs used to follow the same structure but the installation instructions have diverged depending on what deployment options have been available at the time new plugins have been introduced.

mythi avatar Jul 20 '22 17:07 mythi