now hiveserver2 has multiple instance register to zookeeper, if zookeeper recover from fault, the znode represent the hs2 instance got deleted(eg. session timeout), the hiveserver2 will receive NodeDeleted event, in this event, this hiveserver instance will unregister from zookeeper, suppose all hiveserver instance do like this, then from end user perspective ,the hiveserver cluster is no working.
so as per my opinion, the DeRegisterWatcher feature is not required. how do you think ? ________________________________ wenli Regards
