dev: Make setup-single-cluster capable of setting up multiple clusters

[p-se]

By using different git worktrees with different environment variables for configuring the single-cluster environment, specifically the FLEET_E2E_CLUSTER variable, you can use ./dev/setup-single-cluster to start independent k3d clusters next to each other.

Please note that, at the time of creating a cluster, there is a possibility of a conflict still. While clusters can be run at the same time with dev/setup-single-cluster, they cannot be created simultaneously (yet).

• Perhaps add a entry in dev/README.

[weyfonk]

This is a use case I tend to avoid locally for performance reasons, but this may be useful to others.

Would something like this be more accurate?

Suggested change

	# conflicts when the clusters are created simultaneously. It might be worth thinking about
	# how to make the image name unique for each test run.
	# conflicts when multiple Fleet builds are meant to run simultaneously. It might be
	# worth thinking about how to make the image name unique for each test run.

My understanding is that clusters don't need to be created simultaneously, but rather used simultaneously, where an image meant to be used in one could end up being imported in another, just because the same image name and tag are used across all clusters.

As to ways to make image names unique, I suppose we could add an option to override the default dev tag on built Fleet images, eg. with the currently checked out branch, which could prevent conflicts: when using git worktrees, a given branch may only be checked out by a single worktree at a time.

Happy to discuss.

[p-se]

Thank for taking the time to explain to me why this was confusing! Please let me know if the updated phrasing (and position of the comment) is not a sufficient improvement!

[weyfonk]

This simultaneously alone may leave the reader wondering what simultaneity we're talking about here.

Suggested change

	# when fleet is set up simultaneously, as several scripts are calling
	# when fleet is set up simultaneously in multiple sets (ie 1 upstream +
	# 0 to n downstream) of clusters, as several scripts are calling

[weyfonk]

I'm not sure what calling a script "sequentially" means.
What would you think of something like this? How does it compare to what you have/had in mind?

Suggested change

	# be avoided by making sure this script is not called sequentially, e.g. by
	# creating multiple clusters on a single host sequentially rather than
	# simultaneously.
	# be avoided by making sure that images `rancher/fleet:dev` and
	# `rancher/fleet-agent:dev` are rebuilt from the right context (eg. git clone,
	# checkout or worktree) before being imported into clusters created for testing
	# that context.
	# This can be easily achieved by ensuring that, for instance, at most one instance
	# of `./dev/setup-single-cluster` or `./dev/setup-multi-cluster` runs at any given
	# point in time.