• in a high availability cluster how to make sure etcd instances know about each other?