As part of the upgrade process, I enabled HA Namenode which was not previously enabled. I was having issues with one of the namenodes, so I used the Ambari interface to move the namenode to a different host. This is what caused the Hive problems. Hive was pointing to the old namenode. I discovered this by looking in /var/log/hive/hivemetastore.log and discovered there was a timeout trying to connect to the old namenode. I then ran: export HIVE_CONF_DIR=/etc/hive/conf.server and hive –service metatool –listFSRoot. That confirmed hive was configured to point to the wrong namenode. I then updated the configuration via: hive –service metatool -updateLocation <new ha namespace name> <old namenode host info>.
It seems to me this process should have been done as part of the Ambari namenode move process. Is this a known issue? Or is this a case where this should have happened and didn’t for some reason?