RFC - kubectl exit code normalization

[Ricardo Katz]

Hello everybody, hope everything is fine and everyone is safe :)

SIG-CLI decided that it's time to discuss kubectl exit code normalization. There are a number of issues around there from users complaining that there's an error but exit code is zero, or that they need some more information about why the error occurred (was client side? was server side? was a timeout?)

So I've started to write a KEP and would like to request for comments. It's really a WIP, so feel free to agree/disagree on something, propose changes, etc:

https://github.com/kubernetes/enhancements/pull/2574

The idea is to standardize the behavior and document the possible kubectl exit codes.

Because this might have direct impact not only on users, but also on CI/CD providers, automated tasks, as well as people that already rely on the current kubectl behavior and is probably not expecting a change on the command exit code, I kindly ask you to make this proposal reach the maximum interested or impacted people so with yours and their feedbacks we can select a final design, the coding approach and planning, as well as the deprecation path of the old behavior.

Thank you!