If you are going to explicitly set up your cluster seed nodes, if not using ConductR, then you should use the IP addresses of your nodes in the config versus their host names. The same goes for looking up or deploying actors to a remote system via remoting. This advice goes double if you are in the cloud. The reason is that a blip or issue with DNS will cause the failure detector to think a node is not available when it clearly is still up and running. If the DNS problem persists, then the node will eventually get quarantined, and you will need to restart it even if the DNS issue clears.

Using IP addresses of your clustering or remoting config, instead of host...