Currently the boostrap fails when a token is provided by the user on `master init` and works when the token is generated. This is because of a mismatch of how the token string in the kube-discovery secret is encoded.