Detect and log unsupported region error for fully-private clusters

[cPu1]

Description

As discovered in #3133 , eu-south-1 does not support fully-private clusters. This PR detects this error condition and logs the error.

Checklist

• Added tests that cover your change (if possible)
• Added/modified documentation as required (such as the README.md, or the userdocs directory)
• Manually tested
• Made sure the title of the PR is a good description that can go into the release notes
• (Core team) Added labels for change area (e.g. area/nodegroup) and kind (e.g. kind/improvement)

BONUS POINTS checklist: complete for good vibes and maybe prizes?! 🤯

• Backfilled missing tests for code in same general area 🎉
• Refactored something and made the world a better place 🌟

[michaelbeaumont]

What's the advantage over errors.Wrapf?

[cPu1]

It lets the caller identify errors of this type without relying on string comparison.

[michaelbeaumont]

In this case we only print the error anyway, right? I.e. why do we need the errors.As check?

[michaelbeaumont]

Not that it's an issue necessarily but is the handling of the error correct here?

[cPu1]

In this case we only print the error anyway, right? I.e. why do we need the errors.As check?

We could just log the error here instead of wrapping it in a typed error, but I have avoided doing any I/O here because this code has no external side effects (if the provider implementation is mocked). We don't want logs for this case appearing in our unit tests. Additionally, the error type provides a standard format for unsupported features that the calling code can identify with a single check. The errors.As check is required for the caller to identify errors of this type because this check happens higher up in the chain where this error may already be wrapped by other errors.

[michaelbeaumont]

i don't mean logging it here but if you were to just return errors.Wrap(err, fmt.Sprintf("fully-private clusters are not supported in region %q, please retry with a different region", region)) here or even errors.New(fmt.Sprintf("fully-private clusters are not supported in region %q, please retry with a different region", region), wouldn't the user see the same thing they're seeing with this PR? We call logger.Critical in pkg/ctl/create/cluster.go anyway.
Like:

				ufe := &api.UnsupportedFeatureError{}
				if errors.As(err, &ufe) {
					logger.Critical(ufe.Message)
				}
				logger.Critical("%s\n", err.Error())

results in logger.Critical being called for this UnsupportedFeatureError twice basically

[cPu1]

With errors.Wrap or errors.New, the log message will not start with the "fully-private clusters are not supported" error, the messages from the errors.Wrap call higher up in the chain will appear first, which can hide the user-friendly error message.

Having an error type with a Message field to represent this condition lets us log a more-specific, user-friendly message while still preserving the error context and logging it later.

[michaelbeaumont]

Ah the UnsupportedFeatureError itself is being wrapped again and isn't the highest level error? Now I get it.
Sorry, I somehow was convinced the logging happened directly after getting the UnsupportedFeatureError.

[cPu1]

Yes, it isn't the highest-level error.