3.1.2 Release Note

[morrySnow]

New Features

Storage and Compression

• Configurable table compression type — allows selecting specific compression algorithms per table.
  #56276
• Adaptive compaction write caching — dynamically tunes write caching during base compaction rowset flush.
  #56278

Cloud and Object Storage

• Cloud mode query freshness control — adds user-defined tolerance between data latency and consistency.
  #56390
• Object storage endpoint validation relaxed — enables private or custom storage endpoints.
  #56641

Datalake

• OSS support for Datalake VPC Endpoints (dlf/datalake-vpc).
  #56476
• AWS Glue Catalog now supports accessing S3 via IAM AssumeRole.
  #57036
• S3 Client updated to use CustomAwsCredentialsProviderChain for improved credential management.
  #56943

Functional Enhancements

• Java UDF now supports IP type.
  #56346
• BE REST API adds RunningTasks output for monitoring.
  #56781
• Transaction monitoring adds BRPC write-amplification metrics.
  #56832

---

Optimizations

Query Execution and Planner

• COUNT(*) optimization — automatically selects the smallest column to reduce scan load.
  #56483
• Compaction skips empty rowsets to improve throughput.
  #56768
• Warmup statistics add “skipped rowset” metric for better visibility.
  #56373

Storage Layer

• Variant column cache added for sparse columns to speed up reads.
  #56730
• Segment footer is now cached in Index Page Cache to reduce latency.
  #56459
• Recycler supports parallel cleanup tasks to increase throughput.
  #56573

Datalake

• Paimon Time Travel improved and schema mismatch fixed.
  #56338
• Iceberg scan error messages refined and nested namespaces supported.
  #56370, #57035
• Legacy DLF catalog properties removed.
  #56196, #56505
• JSON Load now defaults to row-by-row parsing mode for line-based data.
  #56736

---

Bug Fixes

Datalake

• Fixed Iceberg system table classloader error.
  #56220
• Fixed Iceberg partition table failure when no partition values exist.
  #57043
• Fixed S3A catalog not using IAM AssumeRole profile properly.
  #56250
• Disabled Hadoop FileSystem cache for multi-config object storage catalogs.
  #57153

Query Execution and SQL Engine

• Fixed COUNT pushdown logic error.
  #56482
• Fixed UNION local shuffle behavior bug.
  #56556
• Fixed crash in IN predicate for OLAP storage types.
  #56834
• Fixed timestampdiff computation error for datetimev1.
  #56893
• Fixed crash caused by explode() function.
  #57002

Storage and Load

• Fixed S3 Load check failure when no source file exists.
  #56376
• Fixed FileCache cleanup crash.
  #56584
• Fixed delete bitmap not cleared in MOW compression.
  #56785
• Fixed Outfile bz2 compression failure for small files.
  #57041

Cloud and Recycler Mechanism

• Fixed Warmup skipping multi-segment rowsets.
  #56680
• Fixed CloudTablet Warmup coredump on reference capture.
  #56627
• Fixed Recycler null pointer crash in cleanup task.
  #56773
• Fixed uncaught partition boundary error in Cloud mode.
  #56968

System and Miscellaneous

• Fixed incorrect Prometheus metric format in FE.
  #57082
• Fixed auto-increment value incorrect after FE restart.
  #57118
• Fixed SHOW CREATE VIEW missing column definitions.
  #57045
• Fixed HDFS Reader crash when sampling Profile data.
  #56950

---

Contributors

@0AyanamiRei @924060929 @CalvinKirs @Gabriel39 @JNSimba @MoanasDaddyXu @Mryange @SWJTU-ZhangLei @TangSiyang2001 @Yukang-Lian @amorynan @bobhan1 @csun5285 @dataroaring @deardeng @eldenmoon @englefly @freemandealer @gavinchou @hubgeter @jacktengg @kaijchen @koarz @liaoxin01 @luwei16 @morningman @morrySnow @mrhhsg @mymeiyi @seawinde @shuke987 @sollhui @starocean999 @suxiaogang223 @wyxxxcat @xy720 @yujun777 @zclllyybb @zddr @zgxme @zy-kkk @zzzxl1993