Kubeadm: Create `[Feature:kubeadm]` e2e tests

[luxas]

...under test/e2e_kubeadm/ in the core kubernetes repo.
e2e tests here should for instance make sure that:

• At least one node is tainted/labelled with node-role.kubernetes.io/master=""
• cluster-info is exposed in a correct manner, and the right Role and RoleBinding exist
• The right Node Bootstrap Token RBAC rules exist, the SAR approver is enabled, etc.
• The kubeadm-config ConfigMap exists.
• API Aggregation features are enabled
• What else?

@timothysc @pipejakob Ideas?

[timothysc]

So this requires segregating into it's own location for future forking.

[timothysc]

/assign @fabriziopandini @liztio

This will be a requirement for some of the HA work in order to validate it.

[liztio]

Can we clarify this a little more? What RBAC roles should be expected? What's the "correct manner" for cluster-info to be exposed? What node bootstrap token rules should exist?

[fabriziopandini]

@liztio I think that this document can be a good starting point
https://kubernetes.io/docs/reference/setup-tools/kubeadm/implementation-details/#wait-for-the-control-plane-to-come-up.
If I got this right we have to check that kubeadm-- executes properly the steps starting from mark master.

Is there a WIP for this work?

[liztio]

@fabriziopandini my WIP is https://github.com/kubernetes/kubernetes/pull/62665.

I'm feeling a bit weird about this though... most of these things aren't really E2E tests, they're more like integration tests.

[timothysc]

@liz I'm totally ok with getting what you have hooked up, and we can create a new tracking issue with a list. You could easily back-hole on this, and we could federate the work items across the SIG for this type of work.

[liztio]

Also added https://github.com/kubernetes/test-infra/pull/7833

[luxas]

Reopening until we've got these running automatically

[timothysc]

Per discussion, I think we should separate out the automation from the enablement so move to close.

[liztio]

Closed in favor of #813