Fix proposed with lp:~cgb-cs/appscale/tools-with-zero-db-detection - the ips.yaml above has only a single database node, which means it gets assigned as db_master. The tools were only checking for database nodes (of which db_master is not), so it counted a total of zero database nodes (since database means db_slave) and thus assigned replication = 0, which caused the unavailable exception thrown by Cassandra.
Fix committed causes tools to die if they detect zero database nodes and properly counts database masters as database nodes.
Fix proposed with lp:~cgb-cs/appscale/tools-with-zero-db-detection - the ips.yaml above has only a single database node, which means it gets assigned as db_master. The tools were only checking for database nodes (of which db_master is not), so it counted a total of zero database nodes (since database means db_slave) and thus assigned replication = 0, which caused the unavailable exception thrown by Cassandra.
Fix committed causes tools to die if they detect zero database nodes and properly counts database masters as database nodes.