When Hawkular Metrics deploys and when there are schema changes to apply, start up may fail if any of the Cassandra nodes is down (assume a multi node cluster). This is because the replication_factor defaults to 1, and Hawkular Metrics does not change it. If a node is down, and a request is targeted for that node, then the request will fail even if all other Cassandra nodes are up. While we want to support replication, it requires some management functionality (namely running scheduled repairs) that we do not yet have in place. For now, I propose we do not proceed with initialization and schema updates until all Cassandra nodes are up.