[proposal] A windows k8s Swiss army knife as a utility in the carvel family

[carvel-dev]

[proposal] A windows k8s Swiss army knife as a utility in the carvel family

[cppforlife]

@jayunit100 i dont think proposed functionality would fit well into carvel at this point since we are curating general purpose tools that also have fairly stable API boundaries:

Building windows images and uploading them to various clouds as ISOs.

tools like packer already allow you to do this (if packer didnt exist i could potentially see building a tool like that but it wouldnt necessarily be specific to windows). it's not clear what we would add here.

Bolting 'windows' nodes on to a CluterAPI cluster (wrapper to clusterctl/ powered by ytt templates)

this doesnt feel like an independent tool but more of a plugin to clusterctl (if not something that ought to be handled to by clusterctl directly since for users of clusterapi, linux vs windows machine mgmt should have same UX).

Validating a windows cluster with a few simple smoke tests (wrapper to e2e.test)

this functionality may be more fitting as a plugin to https://github.com/vmware-tanzu/sonobuoy/

[jayunit100]

We have, some rough prototyping in place for this as is, but happy to share more if anyone wants to catch up some time. Also happy to close if this doesn't fit in with the overall carvel model.

My reason for thinking this is perfect for carvel is that, windows clusters are:

• highly heterogenous
• require complex update machinery, so they can't be stamped with robot automation easily

The thus really benefit alot from modular tool that can be mixed and matched into different various downstream workflows.

[cppforlife]

@jayunit100 if im reading this right, this tool seems to be angled towards supporting windows for clusterapi. i have a feeling that this may be a better fit under clusterapi umbrella?

[jayunit100]

it's not capi specific at all , but yes I do believe some capi awareness will be important in it.

Also... I don't think going through kubernetes-sigs/ is worth it as it's not going to be benefited/driven by a heavyweight process .

But if it's not relevant we can just do it on a personal GitHub .

[jayunit100]

closing

[microwavables]

moved from /carvel-community to /carvel as we are merging the two into one