AnsweredAssumed Answered

Documentation around hivemetastore and hiveserver2

Question asked by mandoskippy on Jan 11, 2014
Latest reply on Jan 12, 2014 by Ted Dunning
I am setting up a new test cluster (yea for testing!) and decided it would be a good time dive in to hiveserver2 and hivemetastore services. I am following instructions from http://doc.mapr.com/display/MapR/Hive#Hive-Prerequisites and have some questions. 

Ok, from the docs:

Installing Hive, HiveServer2, and Hive Metastore
The following procedures use the operating system package managers to download and install Hive from the MapR Repository. If you want to install this component manually from packages files, see Packages and Dependencies for MapR Software.
As of MapR version 3.0.2 and ecosystem release 1310 (mapr-hive-*-1310), Hive is distributed as three packages:
mapr-hive - contains the following components:
The core Hive package.
HiveServer2 - allows multiple concurrent connections to the Hive server over a network.
Hive Metastore - stores the metadata for Hive tables and partitions in a relational database.
mapr-hiveserver2 - allows HiveServer2 to be managed by the warden, allowing you to start and stop HiveServer2 using maprcli or the MapR Control System. The mapr-hive package is a dependency and will be installed if you install mapr-hiveserver2. At installation time, Hiveserver2 is started automatically.
mapr-hivemetastore - allows Hive Metastore to be managed by the warden, allowing you to start and stop Hive Metastore using maprcli or the MapR Control System. The mapr-hive package is a dependency and will be installed if you install mapr-hivemetastore. At installation time, the Hive Metastore is started automatically.
This procedure is to be performed on a MapR cluster (see the Installation Guide) or client (see Setting Up the Client).

One of the neat things about hivemetastore and hiveserver2 is having services be managed by mapr.  Since I am running everything in a resourced limited environment, I choose to use a client.  The last sentence there, i.e. this procedure is to be performed on a cluster OR client I thought meant things can be done either way.

Well, it seams that after I have set everything up on a cluster here, that since I choose the client, I realized now that there is no warden, and thus it's not really a good use of the word or, i.e. if you want managed services, then you really have to do a cluster node.  Is this statement true or am I missing something?

Outcomes