Kask currently uses a single hostname as the Cassandra cluster contact point. Once a connection is established, all of other contact points are learned, but if the node listed is down, startup could be delayed. We should convert this into a list of contacts, which will allow us to template the entire node list in Puppet.