In this chapter, we learned that we should run more than one instance of HMaster for high availability and whether we can run two to three HMasters per rack if the cluster is big enough. We should run separate instances of five to seven ZooKeepers on separate hardware machines in a production cluster. Some ZooKeepers can be cohosted (on a production cluster, ZooKeeper machines must be hosted separately). We should keep the time synchronized between all the nodes in an HBase cluster. Run HBasehbck, which checks HBase and tells us if there are any errors in HBase; also, we should make it as a cron job to see the status of the cluster. Run the HBase balancer with data throughout the region server; how to do this will be discussed in Chapter 6, HBase Cluster Maintenance and Troubleshooting.

We will discuss more about HBase's data types, how data is stored in HBase (Logical View/Actual Physical view), and services such as table, row, column family, column, and cells that HBase offers in...