Currently Spark's master is accessible via its public address and slave list contains worker's public IP addresses, so hosts talk to each other via public interfaces. On SL private interfaces offer better performance, so it is necessary to provide an option to use private addresses for clustering.
Currently Spark's master is accessible via its public address and slave list contains worker's public IP addresses, so hosts talk to each other via public interfaces. On SL private interfaces offer better performance, so it is necessary to provide an option to use private addresses for clustering.