Release 368 (11 Jan 2022)#
General#
- Allow setting per task memory limits via - query.max-total-memory-per-taskconfig property or via- query_max_total_memory_per_tasksession property. (#10308)
- Improve wall time for query processing with the - phasedscheduling policy. The previous behavior can be restored by setting the- query.execution-policyconfiguration property to- legacy-phased. (#10350)
- Enable - phasedscheduling policy by default. The previous behavior can be restored by setting the- query.execution-policyconfiguration property to- all-at-once. (#10455)
- Improve performance of arithmetic operations involving decimals with precision larger than 18. (#10051) 
- Reduce risk of out-of-memory failure on congested clusters with high memory usage. (#10475) 
- Fix queries not being unblocked when placed in reserved memory pool. (#10475) 
- Prevent execution of - REFRESH MATERIALIZED VIEWfrom getting stuck. (#10360)
- Fix double reporting of scheduled time for scan operators in - EXPLAIN ANALYZE. (#10472)
- Fix issue where the length of log file names grow indefinitely upon log rotation. (#10394) 
Hive connector#
- Improve performance of decoding decimal values with precision larger than 18 in ORC, Parquet and RCFile data. (#10051) 
- Disallow querying the properties system table for Delta Lake tables, since Delta Lake tables are not supported. This fixes the previous behavior of silently returning incorrect values. (#10447) 
- Reduce risk of worker out-of-memory exception when scanning ORC files. (#9949) 
Iceberg connector#
SingleStore (MemSQL) connector#
- Add support for - timetype. (#10332)
Oracle connector#
- Fix incorrect result when a - datevalue is older than or equal to- 1582-10-14. (#10380)
Phoenix connector#
- Add support for reading - binarytype. (#10539)
PostgreSQL connector#
- Add support for accessing tables created with declarative partitioning in PostgreSQL. (#10400)