4.0.0-RC02

4.0.0-RC02

Release Date: September 29, 2025

New Features

• Supports setting sort keys when creating Iceberg tables.
• Supports Multi-Table Write-Write Transactions, allowing users to atomically commit INSERT, UPDATE, and DELETE operations. These transactions are compatible with both Stream Load and INSERT INTO interfaces, ensuring cross-table consistency in ETL and real-time ingestion scenarios.
• Supports modifying aggregation keys of aggregate tables.

Improvements

• Optimized Delta Lake Catalog cache configuration: adjusted default values of DELTA_LAKE_JSON_META_CACHE_TTL and DELTA_LAKE_CHECKPOINT_META_CACHE_TTL to 24 hours, and simplified Parquet handler logic. #63441
• Improved Delta Lake Catalog error log format and content for better debugging. #63389
• External groups (e.g., LDAP Group) now support role grant/revoke and display, improving SQL syntax and test coverage for stronger access control. #63385
• Strengthened Stream Load parameter consistency checks to reduce risks caused by parameter drift. #63347
• Optimized Stream Load label passing mechanism to reduce dependencies. #63334
• Improved ANALYZE PROFILE format: ExplainAnalyzer now supports grouping metrics by operator. #63326
• Enhanced QueryDetailActionV2 and QueryProfileActionV2 APIs to return results in JSON format. #63235
• Improved predicate parsing in scenarios with large numbers of CompoundPredicates. #63139
• Adjusted certain FE metrics to be leader-aware. #63004
• Enhanced SHOW PROCESS LIST with Catalog and Query ID information. #62552
• Improved BE JVM memory monitoring metrics. #62210
• Optimized materialized view rewrite logic and log outputs. #62985
• Optimized random bucketing strategy. #63168
• Supports resetting AUTO_INCREMENT start value with ALTER TABLE <table_name> AUTO_INCREMENT = 10000;. #62767
• Group Provider now supports matching groups by DN. #62711

Bug Fixes

The following issues have been fixed:

• Incomplete Left Join results caused by ARRAY low-cardinality optimization. #63419
• Incorrect execution plan generated after materialized view aggregate pushdown rewrite. #63060
• Redundant warning logs printed in JSON field pruning scenarios when schema fields were not found. #63414
• Infinite loop caused by SIMD Batch parameter errors when inserting DECIMAL256 data in ARM environments. #63406
• Three storage-related issues: #63398
  • Cache exception when disk path is empty.
  • Incorrect Azure cache key prefix.
  • S3 multipart upload failure.
• ZoneMap filter invalidation after CHAR-to-VARCHAR schema change with Fast Schema Evolution. #63377
• ARRAY aggregation type analysis error caused by intermediate type ARRAY<NULL_TYPE>. #63371
• Metadata inconsistency in partial updates based on auto-increment columns. #63370
• Metadata inconsistency when deleting tablets or querying concurrently. #63291
• Failure to create spill directory during Iceberg table writes. #63278
• Ranger Hive Service permission changes not taking effect. #63251
• Group Provider did not support IF NOT EXISTS and IF EXISTS clauses. #63248
• Errors caused by using reserved keywords in Iceberg partitions. #63243
• Prometheus metric format issue. #62742
• Version check failure when starting replication transactions with Compaction enabled. #62663
• Missing Compaction Profile when File Bunding was enabled. #62638
• Issues handling redundant replicas after Clone. #62542
• Delta Lake tables failed to find partition columns. #62953
• Materialized views did not support Colocation in shared-data clusters. #62941
• Issues reading NULL partitions in Iceberg tables. #62934
• SQL syntax error caused by single quotes in Histogram statistics MCV (Most Common Values). #62853
• KILL ANALYZE command not working. #62842
• Failure collecting Stream Load profiles. #62802
• Incorrect CTE reuse plan extraction. #62784
• Rebalance failure due to incorrect BE selection. #62776
• User Property priority is lower than Session Variable. #63173