In Hadoop 1, the version history was somewhat convoluted with multiple forked branches in the 0.2x range, leading to odd situations, where a 1.x version could, in some situations, have fewer features than a 0.23 release. In the version 2 codebase, this is fortunately much more straightforward, but it's important to clarify exactly which version we will use in this book.

Hadoop 2.0 was released in alpha and beta versions, and along the way, several incompatible changes were introduced. There was, in particular, a major API stabilization effort between the beta and final release stages.

Hadoop 2.2.0 was the first general availability (GA) release of the Hadoop 2 codebase, and its interfaces are now declared stable and forward compatible. We will therefore use the 2.2 product and interfaces in this book. Though the principles will be usable on a 2.0 beta, in particular, there will be API incompatibilities in the beta. This is particularly important as MapReduce v2 was back-ported to Hadoop 1 by several distribution vendors, but these products were based on the beta and not the GA APIs. If you are using such a product, then you will encounter these incompatible changes. It is recommended that a release based upon Hadoop 2.2 or later is used for both the development and the production deployments of any Hadoop 2 workloads.