Apache Doris 3.0.6 Released

Dear community members, the Apache Doris 3.0.6 version was officially released on Jun 16, 2025.

Quick Download: https://doris.apache.org/download/

GitHub Release: https://github.com/apache/doris/releases

Behavior Changes

• Prohibited time-series compaction for Unique tables #49905
• Adjusted Auto Bucket size to 10GB per bucket in compute-storage separation scenarios #50566

New Features

Lakehouse

• Added support for accessing Iceberg table formats in AWS S3 Table Buckets
  • For detailed information, please refer to documentation: Iceberg on S3 Tables

Storage

• IAM Role authorization support for object storage access Applies to import/export, backup/restore, and compute-storage separation scenarios #50252 #50682 #49541 #49565 #49422
  • For detailed information, please refer to documentation

New Functions

• json_extract_no_quotes
  • For detailed information, please refer to documentation
• unhex_null
  • For detailed information, please refer to documentation
• xpath_string
  • For detailed information, please refer to documentation
• str_to_map
• For detailed information, please refer to documentation
• months_between
  • For detailed information, please refer to documentation
• next_day
  • For detailed information, please refer to documentation
• format_round:
  • For detailed information, please refer to documentation

Improvements

Storage

• Streamlined Compaction Profile and logs #50950
• Enhanced scheduling strategy to improve Compaction throughput #49882 #48759 #51482 #50672 #49953 #50819
• Reduced redundant log output #51093
• Implemented blacklist mechanism to prevent Routine Load from distributing metadata to unavailable BEs #50587
• Increased default value of load_task_high_priority_threshold_second #50478

Storage-Compute Decoupled

• Startup optimization: Accelerated File Cache initialization #50726
• Query acceleration: Improved File Cache performance #50275 #50387 #50555
• Metadata optimization: Resolved performance bottlenecks caused by get_version #51111 #50439
• Garbage collection acceleration: Improved object reclamation efficiency #50037 #50766
• Stability enhancement: Optimized object storage retry strategy #50957
• Granular profiling: Added tablet/segment footer dimension metrics #49945 #50564 #50326
• Schema Change resilience: Enabled New Tablet Compaction by default to prevent -230 errors #51070

Lakehouse

Catalog enhancements

• Added partition cache TTL control (partition.cache.ttl-second) for Hive Catalog #50724
  • For detailed information, please refer to documentation: Metadata Cache
• Supported skip.header.line.count property for Hive tables #49929
• Added compatibility for Hive tables using org.openx.data.jsonserde.JsonSerDe #49958
  • For detailed information, please refer to documentation: Text Format
• Upgraded Paimon to v1.0.1
• Upgraded Iceberg to v1.6.1

Functional extensions

• Added support for Alibaba Cloud OSS-HDFS Root Policy #50678
• Dialect compatibility: Returned query results in Hive format #49931
  • For detailed information, please refer to documentation: SQL Convertor

Asynchronous Materialized Views

• Memory optimization: Reduced memory footprint during transparent rewriting #48887

Query Optimizer

• Improved bucket pruning performance #49388
• Enhanced lambda expressions: Supported external slot references #44365

Query Execution

• TopN query acceleration: Optimized performance in compute-storage separation scenarios #50803
• Function extension: Added variable parameter support for substring_index #50149
• Geospatial functions: Added ST_CONTAINS, ST_INTERSECTS, ST_TOUCHES, and ST_DISJOINT #49665

Core Components

• Memory tracker optimization: ~10% performance gain in high-concurrency scenarios #50462
• Audit log enhancement: Added audit_plugin_max_insert_stmt_length to limit INSERT statement length #51314
  • For detailed information, please refer to documentation: Audit Plugin
• SQL converter control: Introduced session variables sql_convertor_config and enable_sql_convertor_features
  • For detailed information, please refer to documentation: SQL Dialect

Bug Fixes

Data Ingestion

• Fixed transaction cleanup failures in BE #50103
• Improved error reporting accuracy for Routine Load #51078
• Prevented metadata task distribution to disable_load=true nodes #50421
• Fixed consumption progress rollback after FE restart #50221
• Resolved Group Commit and Schema Change conflict causing Core Dump #51144
• Fixed S3 Load errors when using HTTPS protocol #51246 #51529

Primary Key Model

• Fixed duplicate key issues caused by race conditions #50019 #50051 #50106 #50417 #50847 #50974

Storage

• Fixed CCR and disk balancing race conditions #50663
• Corrected missing persistence of default partition keys #50489
• Added Rollup table support for CCR #50337
• Fixed edge case when cooldown_ttl=0 #50830
• Resolved data loss caused by GC and Publish contention #50343
• Fixed partition pruning failure in Delete Job #50674

Storage-Compute Decoupled

• Fixed Schema Change blocking Compaction #50908
• Solved object reclamation failure when storage_vault_prefix is empty #50352
• Resolved query performance issues caused by Tablet Cache #51193 #49420
• Eliminated performance jitter from residual Tablet Cache #50200

Lakehouse

• Export fixes
  • Fixed FE memory leaks #51171
  • Prevented FE deadlocks #50088
• Catalog fixes
  • Enabled composite predicate pushdown for JDBC Catalog #50542
  • Fixed Deletion Vector reading for Alibaba Cloud OSS Paimon tables #49645
  • Supported comma-containing Hive partition values #49382
  • Corrected MaxCompute Timestamp column parsing #49600
  • Enabled information_schema system tables for Trino Catalog #49912
• File formats
  • Fixed LZO compression reading failures #49538
  • Added legacy ORC file compatibility #50358
  • Corrected complex type parsing in ORC files #50136

Asynchronous Materialized Views

• Fixed refresh miss when specifying both start time and immediate trigger modes #50624

Query Optimizer

• Fixed rewriting errors with lambda expressions #49166
• Resolved planning failures with constant group by keys #49473
• Corrected constant folding logic #50142 #50810
• Completed system table information retrieval #50721
• Fixed column type handling when creating views with NULL Literal #49881

Query Execution

• Fixed BE crashes caused by illegal JSON values during import #50978
• Corrected Intersect results with NULL constant inputs #50951
• Fixed predicate mis-execution with Variant types #50934
• Resolved get_json_string errors with illegal JSON Paths #50859
• Aligned function behaviors (JSON_REPLACE/INSERT/SET/ARRAY) with MySQL #50308
• Fixed array_map crashes with empty parameters #50201
• Prevented core dumps during abnormal Variant-to-JSONB conversion #50180
• Added missing explode_json_array_json_outer function #50164
• Aligned results between percentile and percentile_array #49351
• Optimized UTF8 handling for functions (url_encode/strright/append_trailing_char_if_absent) #49127

Others

• Fixed audit log loss in high-concurrency scenarios #50357
• Prevented metadata replay failures during dynamic partition creation #49569
• Solved Global UDF loss after restart #50279
• Aligned View metadata return format with MySQL #51058