Within Hadoop, we will mostly talk about HDFS when discussing filesystems and data storage. But, inside almost all Hadoop 2 installations, there is another service that looks somewhat like a filesystem, but which provides significant capability crucial to the proper functioning of distributed systems. This service is Apache ZooKeeper (http://zookeeper.apache.org) and, as it is a key part of the implementation of HDFS HA, we will introduce it in this chapter. It is, however, also used by multiple other Hadoop components and related projects, so we will touch on it several more times throughout the book.

ZooKeeper started out as a subcomponent of HBase and was used to enable several operational capabilities of the service. When any complex distributed system is built, there are a series of activities that are almost always required and which are always difficult to get right. These activities include things such as handling shared locks, detecting...