Say you have 4 nodes n1, n2, n3 and n4.
You can install hadoop and hbase in distributed mode.
If you are using Hadoop 1.x -
n1 - hadoop master[Namenode and Jobtracker]
n2, n3 and n3 - hadoop slaves [datanodes and tasktrackers]
For HBase, you can choose n1 or any other node as Master node, Since
Master node are usually not CPU/Memory intensive, all Masters can be
deployed on single node on test setup, However in Production its good to
have each Master deployment on a separate node.
Lets say n2 - HBase Master, remaining 3 nodes can act as
Hive and Nutch can reside on any node.
Hope this helps; For a test setup this should be good to go.
For Hadoop 2.x, since your cluster size is small, Namenode HA deployment
can be skipped.
Namenode HA would require two nodes one each for an active and standby
A zookeeper quorum which again requires odd number of nodes so a minimum
of three nodes would be required.
A journal quorum again require a minimum of 3 nodes.
But for a cluster this small HA might not be a major concern. So you can
n1 - namenode
n2 - ResouceManager or Yarn
and remaining nodes can act as datanodes, try not to deploy anything
else on the yarn node.
Rest of the deployment for HBase, Hive and Nutch would remain same.