See issues on GitHub

Release Notes

Upgrade Notes:

  • This is a release candidate release. We recommend testing this version out in developement, and not running it in production yet.
  • Upgrading from 0.20 requires a full cluster restart.

Breaking changes:

  • GeoShape precision (#2803)
  • Remove JMX connector creation flags, and JMX attributes (#2728)
  • Change field data stats header from field_data to fielddata. (#2727)
  • Change the geo_shape support to only rely on the lucene geo_shape support (#2720)

New features:

  • Suggest Feature should have a dedicated REST endpoint (#2774)
  • Implement search shards API (#2726)
  • Add Phrase Suggester (#2709)

Enhancements:

  • Specialise the default codec to reuse Lucene41 files in the default case (#2799)
  • change geo_shape default configuration (#2756)
  • Add KeywordRepeatFilter from Lucene to Beta2 (#2753)
  • Move Smoothing Model into its own sub-object in the Phrase suggest request (#2735)
  • _count should take the preference parameter for targeting specific shards (#2698)
  • “-4” parsed as a date (#2694)
  • Field Data:
    • optimize long type to use narrowest possible type automatically (#2795)
    • Add node level cache type (#2722)
  • Mapping: dynamic flag is explicitly returned even when not set (#2789)
  • Network: A closed channel might not always fire up a close event (#2733)
  • Query DSL: Filtered query to make query optional (defaults to mach_all) (#2718)

Bug fixes:

  • Sort Fails with AIOOB exception if field has rarely a value. (#2773)
  • Repeated ConnectExceptions in logs until node is restarted (#2766)
  • PhraseSuggest CandidateGenerator doesn’t respect size parameter (#2752)
  • has_child query AVG score mode does not always work correctly (#2750)
  • Term/Terms filters on numeric fields gives wrong result (#2746)
  • has_child returns parent and child (#2744)
  • Problems with range searches for time with lte (#2731)
  • Phrase Suggester “size” should override “shard_size” when searching on index with 1 shard (#2729)
  • Interchanged values in field_data stats (#2724)
  • NullPointerException when applying a sort and using ignoreMapped(true) and the field does not exist (0.20.5) (#2711)
  • Nested sort not calculating avg/sum (#2701)
  • Ids filter without a type throws IndexOutOfBounds (#2695)
  • malformed elasticsearch.yml causes unresponsive hang (#2693)
  • Using Java node client and deleting all indexes cause system hungs (#2692)
  • NPE with Fuzzy Like This on non existing field (#2690)
  • Bug when searching concrete and routing aliased indices (#2682)
  • Index dynamic settings might not be allowed to be applied on master with no data node (#2675)
  • Analyze API returns in YAML format if analyzed string begins with - (#2624)
  • null pointer exception in 19.8. same code path exists in master branch (#2546)
  • The _id path should not allow arrays (#2275)
  • Allow index: “no” for _type (#2696)