I'm not a zookeeper expert, but zookeeper is supposed to be characteristics of light-weight, high performance, and fast response. Unless you zookeeper is already overloaded, I don't see why you would need a separate zookeeper cluster just for Hive.
There are a few zookeeper usages in Hive, the additional stress on zookeeper is determined by the load on your HS2. As most of the time user sessions are waiting on query execution, I don't expect the additional stress on your zookeeper will be significant. You do need to test it out before putting it in production as a general practice. On Fri, Apr 17, 2015 at 1:56 PM, Eduardo Ferreira <eafon...@gmail.com> wrote: > Hi there, > > I read on the Hive installation documentation that we need to have a > ZooKeeper cluster setup to support Table Lock Manager (Cloudera docs link > below). > > As we have HBase with a ZooKeeper cluster already, my question is if we > can use the same ZK cluster for Hive. > Is that recommended? > What kind of load and constrains would this put on the HBase ZK cluster? > > > http://www.cloudera.com/content/cloudera/en/documentation/core/latest/topics/cdh_ig_hiveserver2_configure.html > > Thanks in advance. > Eduardo. >
