Release notes

Upgraded to Elasticsearch 8.12.2

Improved the circuit-breaker mechanism for multi-conditional index joins. The circuit-breaker now triggers a failure when the limit of index lookup operations is exceeded at run-time, rather than relying on estimations done at the query planning phase.

Enhanced the failure message of the circuit-breaker mechanism for multi-conditional index joins, incorporating details on the reached limit and the count of remaining lookup operations.

Introduced an option to deactivate the circuit-breaker mechanism for multi-conditional index joins. This can be achieved by setting -1 in the federate.planner.index_join.max_lookups configuration setting.

Fixed a bug concerning missing bounds and multi-value columns.

Introduced an allow_missing_bound parameter in the lt, lte, gt, and gte operators to accept a missing or null value in a field used in the join.

Upgraded to Elasticsearch 8.12.1

Added the truncation trace in search requests executed with the adaptive query planner.

Improved the performance of the within operator by doing a merge overlap optimization.

Implemented a pre-filtering technique in spatial joins to reduce unnecessary index lookup operations.

Enhanced the data type compatibility check for fields used in a join.

Introduced a new resource management layer in the adaptive query planner.

Improved the performance of the index fingerprint computation.

Removed the community license. An empty license is returned if no license is registered.

Added the planner time, execution time, and waiting time in the JSON response.

Improved the examples using aggregation in the user documentation.

Upgraded to Guava 32.1.3.

Upgraded to Java 21.

Fixed a class cast exception from occurring when projecting fields in a multi-conditional index join.

Fixed an issue that created a stage node for a cached join in the adaptive query planner.

Fixed a parsing error message when using an object instead of an array for the indices join query clause.

Fixed an issue with the creation of function_score queries when the boost_mode parameter isn’t passed explicitly.

Fixed an incomplete response when doing an aggregation and no index could match the request due to a date range filter.

Fixed an issue on cache key definition when runtime fields were defined at query time.

Fixed an issue on a cached join when its parent was shared by two other parents.

Fixed an issue when computing the index scan cost, where the field existed but without terms in a segment.

Fixed a concurrency issue in multi-conditional index joins.

Fixed a concurrency issue in Point-In-Time query shard cache.

Fixed a concurrency issue in AQP reports.

Normalized Federate settings. The following planner settings have been marked as deprecated, a new setting is proposed;