Monday, April 14, 2014

Oracle Big Data Appliance node layout

The Oracle Big Data Appliance ships in a full rack configuration with 18 compute nodes that you can use. Each node is a Sun X4-2L or a X3-2L server depending on the fact if you purchased a X4-2 or X3-2 Big data appliance. Both models do however provide you with 18 compute nodes.  The below image shows the rack layout of both the X3-2 and X4-2 rack. Important to note is that the number of servers is done bottom up. A starter rack has only 6 compute nodes and you can expand the rack with in-rack expansions of 6 nodes. Meaning, you can grow from 6 to 12 to a full rack as your capacity needs grow.


In every setup, regardless of the fact if you have a starter rack, a full rack or a rack extended with a single in-rack expansions of 6 nodes (making it a 12 node cluster), 1, 2 and 3 do have a special role. As we go bottom up, starting with node 1 we have the following software running on the nodes:

Node 1:
First NameNode
Zookeeper
Failover controller
Balancer
Puppet master
Puppet agent
NoSQL database
Datanode

Node 2:
Second NameNode
Zookeeper
Failover controller
MySQL backup server
NoSQL Admin
Puppet agent
DataNode

Node 3:
Job Tracker
Zookeeper
CMServer
ODI Agent
MySQL primary server
Hue
Hive
Beeswax
Puppet Agent
NoSQL
DataNode

Node 4 – till 6/12/18
Datanode
Tasktracker
Cloudera manager Agent
Puppet Agent
NoSQL

Understanding what runs where is of vital importance when you are working with an Oracle Big Data appliance. It helps you understand what parts can be brought down without having to much effect on the system and which parts you should be more careful about. As you can see from the above list there are some parts that are made high available and there are some parts that will result in loss of service when brought down for maintenance. 

No comments: