Release Notes

Breaking changes

  • Mapping:
    • Remove unsupported postings_format / doc_values_format #7604 (issues: #7238, #7566)

Enhancements

  • Core:
    • Check if from + size don't cause overflow and fail with a better error #7778 (issue: #7774)
  • Internal:
    • Chunk direct buffer usage by networking layer #7811
    • DistributorDirectory shouldn't search for directory when reading existing file #7323 (issue: #7306)
    • Switch to fixed thread pool by default for management threads #7320 (issue: #7318)
  • Test:
    • use local random instance rather than thread local version #7665

Bug fixes

  • Backport Recovery / Snapshot file identity improvements to 1.3 #7857 (issues: #7351, #7434)
  • Aggregations:
    • Added missing module registration in TransportClient for Significant Terms #7852 (issue: #7840)
    • Inconsistent sorting of top_hits fixed #7697
    • Fixes resize bug in Geo bounds Aggregator #7565 (issue: #7556)
    • The nested aggregator should iterate over the child doc ids in ascending order. #7514 (issue: #7505)
  • Bulk API:
    • Bulk operation can create duplicates on primary relocation #7729
  • Indexed Scripts/Templates:
    • Change the default auto_expand to 0-all #7502
    • Fix .script index template. #7500
  • Internal:
    • Make close() synchronized during node shutdown #7885
    • Ensure GroupShardsIterator is consistent across requests, to ensure consistent sorting #7698
    • Turn unexpected exceptions when reading segments into CorruptedIndexException #7693
    • Unify search context cleanup #7643 (issue: #7625)
    • Use SEARCH threadpool for potentially blocking operations #7624 (issue: #7623)
    • Wait until engine is started up when acquiring searcher #7456 (issue: #7455)
    • Indexes unuseable after upgrade from 0.2 to 1.3 and cluster restart #7430
    • Use node's cluster name as a default for an incoming cluster state who misses it #7414 (issue: #7386)
    • Force optimize was not passed to shard request #7405 (issue: #7404)
    • Ignore segments.gen on metadata snapshots #7379
    • When serializing HttpInfo, return null info if service is not started #6906
  • Java API:
    • Add back string op type to IndexRequest #7387
  • Mapping:
    • Fix index setting in _boost field #7557
    • Keep parameters in mapping for _timestamp and _size even if disabled #7475
  • Nested:
    • If the _type field isn't indexed nested inner docs must be filtered out. #7410
  • Parent/Child:
    • has_child or other p/c query wrapped in a query filter may emit wrong results #7685
    • Add support for the field data loading option to the _parent field. #7402 (issue: #7394)
    • If _parent field points to a non existing parent type, then skip the has_parent query/filter #7362 (issue: #7349)
  • Query DSL:
    • Empty bool {} should return match_all #7347 (issue: #7240)
  • Resiliency:
    • Improve recovery / snapshot restoring file identity handling #7351
  • Scan:
    • Use ConcurrentHashMap in scan search to keep track of the reader states. #7499 (issue: #7478)
  • Snapshot/Restore:
    • Allow to get metadata from arbitrary commit points #7376
    • Fix NPE in SnapshotsService on node shutdown #7322 (issue: #6506)
  • Suggesters:
    • infinite loop in GeolocationContextMapping #7433
    • Bugs with encoding multiple levels of geo precision #7369 (issue: #7368)
  • Transport Client:
    • fixed the node retry mechanism which could fail without trying all the connected nodes #6829