Created 04-18-2017 05:06 AM
1st. Why are you set the NN java heap size to 14GB? I think, it has to be down to 1GB or 2GB. You don't have to assign highly NN java heap size like that.
2rd. Did you check the ntp master & clients time status?
3rd. Check the /var/log/hdfs/hadoop-hdfs-namenode-nn01-FQDN.log. Originally NameNode is recovering from corrupt blocks or missing blocks, and also you'll verify the procedure in the log.
4th. Did you install RegionServer to all datanode servers?