API Server stays in crash-backoff.
Assumes kubernetes master and etcd run co-location on same node.

3 known solutions to discuss:

1. Add etcd headless services/endpoints so api can use etcd by-name via DNS
2. Add a cluster configmap, and map the ENV var to API server, less fault tolerant without changing API server to accept multiple etcd's and perform timeout/rollover.
3. Run etcd as a rc/service dedicated to nds labs API server, unclear of advantages/disadvantages.