NVIDIA / cloud-native-docs

Documentation repository for NVIDIA Cloud Native Technologies
https://docs.nvidia.com/datacenter/cloud-native/
Apache License 2.0
16 stars 18 forks source link

Incomplete table of values for the Helm chart #50

Closed lindhe closed 3 months ago

lindhe commented 5 months ago

I found that the chart values documented here are incomplete. They are at least missing the devicePlugin object as documented here. Would be great if that was updated! 👍

This was originally posted in https://github.com/NVIDIA/gpu-operator/issues/698 and was moved here.

mikemckiernan commented 4 months ago

Hello @lindhe! Thank you for opening this issue. At this point, I don't plan to cover every field in the chart. Can you tell me more about which fields in the device plugin section that you needed to use?

lindhe commented 4 months ago

The particular value I needed to set was devicePlugin.config.name.

I completely get you! It sucks to manually create tables for values and then have the never ending struggle of keeping them up-to-date. If you do plan on having tables at all, I would recommend checking out this: https://github.com/bitnami/readme-generator-for-helm

Otherwise, as long as the values.yaml file is publicly available and easy to find, having tables are just a nice-to-have. Is there an easier way to browse it than running a helm pull on the chart and exploring it locally?

lindhe commented 4 months ago

If the tables will not be complete, I would much appreciate a small note saying something along the lines of:

ℹī¸ Note: this table is not complete. For the complete list of values, please see https://github.com/NVIDIA/some-repo

mikemckiernan commented 3 months ago

@lindhe, thank you for your patience and suggestions.

I spoke with the maintainers of the GPU Operator chart and the consensus was that the most likely use case for setting devicePlugin.config is to configure time slicing. I opened #69 to add the field to the table so it's findable and to add a link to the time slicing procedure.

If you had a different use case, please share here or in the GPU Operator repo and we'll try to address it.

lindhe commented 3 months ago

Time slicing is exactly my use-case. Thanks for the response!