This is a improvement of JIRA
HDFS-10821. In HDFS-10821, it supports the diskbalancer report command can be executed with multiple nodes. But the multiple nodes are read from a nodes string. It will be not good to use when the cluster is big enough. A better way we can do is making that can be read from a host file as that has been done in many places in HDFS(ex. decommission, balancer). This JIRA focus on this point, and will makes this done in diskbalancer.