🌱 Add new management cluster environment option

[yiannistri]

What this PR does / why we need it:

This PR adds a new option (internal-kind) for creating a local bootstrap cluster (using kind with extra port mappings) that accepts requests from the host. This can be useful in environments where the CAPI provider is running in the local network (eg. vSphere/CAPV). The new setup basically replicates this guide.

Which issue(s) this PR fixes (optional, in fixes #<issue number>(, fixes #<issue_number>, ...) format, will close the issue(s) when PR gets merged):
Fixes #1162

Special notes for your reviewer:
I have added a new bootstrap cluster provider instead of updating existing code we have in place for creating a kind cluster since the port mappings settings were ignored when the docker socket extra mount was specified (useful for CAPD tests).

To test the setup of the bootstrap cluster run the following:

MANAGEMENT_CLUSTER_ENVIRONMENT="internal-kind" TAG=v0.0.1 GINKGO_LABEL_FILTER=vsphere make test-e2e

Checklist:

• squashed commits into logical changes
• includes documentation
• adds unit tests
• adds or updates e2e tests

[anmazzotti]

@yiannistri I think the extraMapping are ignored just because they are overwritten by the docker socket code.

Would it be possible to fix this (it's a bug in our test suite anyway) and apply your changes to the current kind environment?

One other question I have is what provides type: LoadBalancer service? Do we have anything installed in the kind cluster for LB?

[yiannistri]

@anmazzotti yes for sure, we can fix this upstream. I have added a new provider in the meantime only to get things working for both CAPD tests as well as the CAPV tests that will need the extra port mappings.

I will create an issue upstream and prepare a PR to address this.

Regarding your second question, I have added a new manifest as part of this PR that creates an nginx load balancer.

[yiannistri]

One downside of this approach (using extra port mappings) is that if the previous kind cluster is still running on the machine, the creation of a new kind cluster will fail because the 443 port will still be allocated to the previous cluster. We can ensure that no kind clusters are running when we begin a new test run, to mitigate this.

[furkatgofurov7]

LGTM thanks for the PR! One question/suggestion inline:

[furkatgofurov7]

@yiannistri thanks for the PR, changes LGTM (we shall take it out of the draft maybe)

[alexander-demicev]

thanks!

[furkatgofurov7]

Thanks!