Deleting a node via Terraform should delete the Kubernetes node

[displague]

When the node count is modified, a destroy triggered provisioned should attempt to drain and delete the node:

for example:

resource "metal_device" "..." {
  // ...
  provisioner "local-exec" {
    when        = destroy
    command = "kubectl -f ${kubeconfig}  delete node ${self.hostname}" // we would want to drain/cordon first. can we get the kubeconfig path in this block?
  }
}

The CCM (#64) would handle the eventual cleanup of deleted Terraform or UI deleted nodes, but this approach will allow for Terraform deleted nodes to be cleaned up less abruptly.

[jmarhee]

I like this idea a lot. This should be doable-- in other locations, we've assumed the kubeadm admin kubeconfig (/etc/kubernetes/admin.conf), so I don't see why we couldn't here.

[jmarhee]

#90 addresses this in a best-effort (makes an attempt, reminds user to clean up manually if KUBECONFIG is not set, because a destroy triggered resource can't consume variables in such a way to allow me to push in a path) using local-exec; perhaps something the Kubernetes provider can address (though if I recall, it cannot delete a node).