This release is mainly to get out dimension compression and rework the druid documentation. There are no update concerns with this version of Druid.

New Features

Added support for Dimension compression of segment columns, enabled by default. Compression is applied to the column storing the dimension value indices, but not to the dimension values themselves. This change only applies to single value dimensions, multi-value dimensions are left uncompressed. With real-world data we have seen segments sizes reduced by 50% for some datasources, but actual compression ratios will vary based on the data. Sparse and repetitive columns will benefit the most, whereas more random and higher cardinality columns will benefit less. Old segments can be converted using the updated VersionConverterTask.
Initial support for Microsoft Azure as a deep storage option has been added. Thanks @davrodpin!

Improvements

Improved VersionConverterTask to allow for an IndexSpec and forced updates. This enables the ability to convert old segments to use dimension compression,
Improved how the datasource metadata query filters on segments to scan.

Ignore rows with invalid interval for index task (#1264)
always re-upload snapshot self-contained jars to hdfs (#1261)
Skip raising false alert when the coordinator loses leadership (#1224)
Fix an issue that after broker forwards GroupByQuery to historical, havingSpec is still applied (#1292). Thanks @guobingkun!
Fix type incorrect types for update sql statement for metadata storage (#1295). Thanks @anubhgup!
fix serde issue when pulling timestamps from cache (#1304)

Reworked the Druid documentation such that it can be consumed in order.
Many documentation fixes and improvements thanks to @bobrik, @infynyxx, @rasahner, @b-slim, @textractor, @truenorth, @gknapp, and others we may have missed!