The Apache Hive PMC has recently voted to release Hive 1.0.0 (formerly known as Hive 0.14.1).
This release is recognition of the work the Apache Hive community has done over the past nine years and is continuing to do. The Apache Hive 1.0.0 release is a codebase that was expected to be released as 0.14.1 but the community felt it was time to move to a 1.x.y release naming structure.
As such, the code changes in 1.0.0 are small in scope. The two main changes are to begin to define the public API and remove HiveServer 1. The API documentation has only started and will be continued in HIVE-9363.
Removal of HiveServer 1 is an important step towards making Hive enterprise-ready. HiveServer 2 was contributed to the Hive project by Cloudera two and half years ago, and has enabled Hive to support JDBC, ODBC, and fined-grain authorization with Apache Sentry (incubating).
For CLI users, migrating to HiveServer2 will require migrating to Beeline. Cloudera and Intel have continued to invest heavily in Beeline and HiveServer2 to make the transition easier. We’ve added too many polish items to cover in any detail. One feature I am particularly excited about is retrieval of query logs via the JDBC API and Beeline query status, which is implemented using that API. This will make it easier for Hive developers to use Beeline to develop their future Hive jobs.
The next major release of Hive, 0.15.0, has been renamed to 1.1.0. The much anticipated Hive-on-Apache Spark work is the marquee feature expected in 1.1.0, and the release process has already started. (In this blog, we’ve covered the Hive-on-Spark project in detail, including the design, first demo, and a hands-on sandbox.)
As always, you can expect Cloudera to pull upstream Hive features into CDH when they are ready for our customers to run in production. Stay tuned.
Brock Noland is a Software Engineer at Cloudera, and a Hive PMC member.