StarRocks version 4.0

warning

Downgrade Notes

• After upgrading StarRocks to v4.0, DO NOT downgrade it directly to v3.5.0 & v3.5.1, otherwise it will cause metadata incompatibility and FE crash. You must downgrade the cluster to v3.5.2 or later to prevent these issues.

• Before downgrading clusters from v4.0.2 to v4.0.1, v4.0.0, and v3.5.2~v3.5.10, execute the following statement:

  SET GLOBAL enable_rewrite_simple_agg_to_meta_scan=false;

  After upgrading the cluster back to v4.0.2 and later, execute the following statement:

  SET GLOBAL enable_rewrite_simple_agg_to_meta_scan=true;

4.0.13

Release Date: July 16, 2026

Behavior Changes

• The escape handling of LIKE predicates with constant operands (folded on the FE) now matches MySQL 8: SELECT 'a\\b' LIKE 'a\\\\b' returns 1 and SELECT 'a\\b' LIKE 'a\\b' returns 0. Queries that relied on the previous non-MySQL escaping semantics will return different results. #74814
• SHOW [FULL] FUNCTIONS now always includes the isolation property (shared or isolated) in the Properties column of UDFs, so users can tell whether the property is set without recreating the function. #75255
• Iceberg REST catalogs with vended credentials use the table metadata cache again, reverting the earlier cache bypass that sent every getTable() to the REST catalog and caused AWS Lake Formation Rate exceeded failures. Cached tables now renew their credentials on every refresh cycle, and the table cache expiry for REST catalogs is additionally capped at 3000 seconds. #75431

Improvements

• Added checksum protection for shared-data tablet metadata and transaction logs. #74924
• Supported combined transaction log / file bundling for FRONTEND_STREAMING loads. #74460
• Scoped shared-data schema-change job locks to the table to reduce lock contention. #75087
• Batch tablet force-delete marking now acquires the TabletInvertedIndex write lock once per batch instead of once per tablet. #75616
• Added an FE metric for the maximum pending-publish time of committed transactions. #75025
• Added a memory limit check for column upgrades in window operator processing. #75821
• Removed unnecessary per-row seeks in the offsets-only read path of array columns. #75861
• Foreground row-count estimation of Iceberg tables is now computed from manifest metadata without enumerating every data file. #75280
• Addressed security vulnerabilities (CVE): excluded the vulnerable org.jline:jline (jline-remote-telnet) from Hadoop transitive dependencies, and upgraded jackson-databind to 2.21.4. #75066 #75373

Bug Fixes

The following issues have been fixed:

• Incremental scan-range scheduling could recompute a different per-driver layout when reusing a deployed fragment instance, leaving part of the scan ranges unconsumed and losing rows (for example, INSERT from Hive). #74674
• INT96 timestamps nested inside ARRAY/MAP/STRUCT in Parquet files read via FILES() or Broker Load missed the session-timezone conversion and were returned shifted by the timezone offset (top-level INT96 columns were correct). #74868
• The audit log did not record the exported row count of SELECT INTO OUTFILE. #74467
• A strict cast could raise an overflow error from the underlying data of NULL rows, which should be ignored. #74903
• parse_json did not respect the ALLOW_THROW_EXCEPTION setting when handling invalid input. #74976
• First-load statistics collection could not be enabled per table while disabled globally: an explicitly set table property now takes precedence over the global configuration enable_statistic_collect_on_first_load. #74794
• Partial column updates on shared-data tables could crash the BE or silently corrupt data when the tablet schema drifted from the transaction schema. #74005
• Unexpected BE process restarts. #74424
• A BE crash (SIGFPE) in Iceberg truncate/bucket partition transforms when the width or bucket count is zero. #74998
• A BE crash caused by a null driver_executor in FragmentContext::set_final_status. #75030
• A race between transaction begin and autovacuum could delete a still-needed transaction log, permanently wedging the partition's publish ("Both txn_log and corresponding tablet_meta missing"). #74906
• avg(DISTINCT x) was incorrectly rewritten to use a sum/count materialized view, returning wrong results. #75071
• A boundary bug in TopN RANK sorting could produce incorrect results. #75045
• split / split_part / str_to_map with an empty delimiter could read out of bounds on invalid UTF-8 input. #75068
• ALTER TABLE ... MODIFY COLUMN ... AFTER a nonexistent column now returns a clear error message. #75073
• A BE crash (SIGFPE) in mod() / pmod() when computing the type's minimum value modulo -1. #74980
• bar() grew memory without bound on a negative or huge width (potential DoS); such inputs are now rejected with an error. #75143
• The transaction-state callback was not unregistered when a multi-statement stream load task was removed. #75188
• A crash when a query was cancelled during spill partition sorting. #75140
• The query memory limit was not enforced during table function execution. #75179
• Selecting a column named floor or ceil failed at parse time with a ClassCastException. #75241
• A heap-use-after-free in OrderedPartitionExchanger when the previous chunk was mutated downstream. #75279
• Three FE metadata-lock correctness races. #74968
• Load spilling could dereference a missing query context when recording spill metrics. #75236
• ADLS2 ListPaths on storage accounts without hierarchical namespace caused CN crashes and vacuum failures. #75166
• BE/CN JVM metrics emitted invalid Prometheus # TYPE lines. #75240
• JIT code generation truncated LARGEINT literals to 64 bits, producing wrong results. #75137
• A combined ALTER TABLE on an external Iceberg table re-executed already-queued actions. #74036
• A nested-loop join crash caused by a build-side column nullability mismatch. #75343
• Iceberg manifest column statistics are now cached selectively to avoid excessive FE memory usage. #75395
• addPhysicalPartition could create only one physical partition per call, making physical-partition backfill of random-distribution tables extremely slow. #75430
• A SQL injection vulnerability in the information_schema.task_runs predicate lookup. #75520
• A BE crash (SIGSEGV) when multiple UNNEST calls in one query referenced the same array column. #75012
• Incorrect translation of nested dictionary-encoded expressions across exchange nodes. #75246
• The join skew hint was lost when the optimizer duplicated expressions. #68964
• Collecting the tables referenced by a view did not skip CTE references. #74813
• A BE abort on CAST(json AS STRUCT<...>) when a struct field name is not a valid JSON path. #75355
• Temporary Parquet dictionary-code columns could leak to upper layers of the execution plan. #74452
• Incorrect results caused by the sort-column elimination optimization. #74983
• The UNNEST result struct type was not narrowed consistently with its input array element type. #75445
• A race between the FE EOS-cancel and the BE stage-2 deployment could mark a successfully finished query as cancelled. #75009
• Join reorder pruning could drop columns that predicates still referenced. #74791
• The global metadata lock was not released when a metadata image dump failed to acquire some database locks, blocking subsequent metadata operations. #75488
• StringSearch::_pattern was left uninitialized, and the split_debug_symbol log output was incorrect. #75614
• LIKE with the single-character wildcard _ returned wrong results on columns with a GIN inverted index. #75551
• The segment-iterator vector could get positionally misaligned during shared-data primary key index rebuild. #74887
• With file bundling enabled, vacuum could delete a bundle file still referenced by sibling tablets because orphaned bundled segments were not flagged as shared; subsequent publishes then failed with "Object ... does not exist". #75689
• histogram() crashed on a non-positive bucket count; it now returns an error. #75041
• A compaction crash (CHECK failure in JsonMergeIterator) when a flat-JSON column changed from NOT NULL to nullable. #75680
• Common predicate operators were lost when the join tuning guide rebuilt a join. #75773
• An UnsupportedOperationException in ApplyTuningGuideRule caused by an immutable inputs list in OptExpression. #70785
• SHOW PARTITIONS and partitions_meta reported the logical-partition bucket count instead of the per-physical-partition count for shared-data tables. #75734
• A memory allocation exception in NLJoinProbeOperator::pull_chunk crashed the BE instead of failing the query. #75788
• SHOW CREATE ROUTINE LOAD emitted a spurious comma before the first load-property clause, making the output non-executable. #75522
• CTAS (CREATE TABLE AS SELECT) did not accept an ENGINE clause, making CTAS into a Unified catalog impossible. #75771
• OR predicates over null-safe-equal (<=>) join conditions were incorrectly rewritten to UNION ALL, returning wrong results. #75038
• A query cache normalization crash for tables with sub-partitions. #75789
• array_map / transform dropped NULL rows when all non-null input arrays were empty. #75141
• regexp_extract_all fell into an infinite loop on a zero-length capture group. #75798
• Flat-JSON subfield reads returned NULL for keys whose name contains ., which could also cause silent row loss when predicates were pushed down. #75583
• UNNEST output struct subfields were pruned by the output's own access group instead of the input array's, which could prune subfields that were still needed. #76002
• The FE Iceberg manifest data-file cache could serve an incomplete file set, silently dropping a data file from scan planning and under-counting query results. #76215
• The Iceberg table metadata cache was not invalidated after an INSERT commit, so subsequent queries could miss the latest snapshot. #67230

4.0.12

Release Date: June 25, 2026

Behavior Changes

• When reading INT64 timestamps from Parquet files written with isAdjustedToUTC=false (timezone-naive), SELECT FROM FILES() and broker/stream LOAD no longer shift the values by the session timezone offset. Such timestamps are now read as wall-clock values, consistent with Trino, Spark, and Impala. Previously the values drifted whenever the session timezone was not UTC. #73674
• CTAS (CREATE TABLE AS SELECT) now preserves the declared VARCHAR(N) length when the source carries an explicit user length (a catalog column reference, CAST AS VARCHAR(N), or a string literal), instead of widening it to VARCHAR(1048576). This keeps the length constraint enforceable and aligns DDL with dbt schema contracts. Materialized view materialization still widens columns as before. #73498
• The Paimon connector now respects the session variable connector_max_split_size when calculating scan splits, instead of always using the default value, so tuning it now affects Paimon scan parallelism. #71756

Improvements

• Optimized base64_to_bitmap by folding the conversion at constant-evaluation time for constant inputs. #74684
• ngram_search now supports a non-constant needle (the search term can be a column expression rather than only a constant). #74675
• The Arrow-to-JSON converter now supports LARGE_LIST and FIXED_SIZE_LIST types. #73714
• Added an opt-in option to isolate wide-string columns during statistics collection to reduce memory pressure. #73258
• information_schema.COLUMNS now populates the DATETIME_PRECISION field. #74623
• Relaxed database read locks to table-scoped intensive locks in InformationSchemaDataSource and FrontendServiceImpl to improve concurrency. #73936 #73913
• Narrowed database write locks to table-scoped intensive write locks for shared-nothing clusters, and scoped replica row-count updates to the table lock. #74523 #74521
• Moved the routine-load broker RPC out of the per-job write lock to reduce contention. #73591
• Deferred JDBC REMARKS fetching out of the getTable() hot path to speed up metadata access for JDBC catalogs. #73488
• Pushed down the table_name predicate for information_schema.tables_config queries. #73210
• Skipped per-replica scans on single-medium BEs in BackendLoadStatistic. #73555
• Added a write timeout to the MySQL channel result send path to prevent stuck connections. #73646
• Added catalog recycle bin size gauge metrics. #74440
• Added vacuum batch-size and retry-count metrics, and added decorrelated jitter to the lake vacuum retry backoff to reduce retry storms. #74112 #74108
• Upgraded third-party dependencies to address security vulnerabilities (CVE): Netty to 4.1.135.Final, Tomcat to 9.0.118, and Thrift to 0.23.0. #74668 #73797 #73625

Bug Fixes

The following issues have been fixed:

• Successfully committed multi-statement transaction stream loads were shown as PREPARING forever in information_schema.loads and SHOW STREAM LOAD. #74386
• Rows were silently dropped from information_schema.loads on clusters whose session timezone differs from Asia/Shanghai, because load times were exchanged as naive wall-clock strings across the BE/FE thrift boundary. #73365
• The COMMIT of an explicit transaction waited only query_timeout milliseconds (instead of seconds) for the database write lock due to a unit mismatch. #73549
• current_timestamp / now() column defaults were displayed as a frozen literal after ALTER TABLE ... ADD COLUMN and could be lost across FE restarts or edit-log replay. #73455
• Querying sys.fe_memory_usage / sys.fe_locks without the OPERATE ON SYSTEM privilege returned a misleading RPC-failure message instead of a clear access-denied error. #73567
• Automatic per-key Hive partition stats refresh could overload the Hive Metastore for tables with many partitions. #73563
• A null-pointer issue when reading the GTID during a schema change. #74855
• An empty analytic operator was not pruned after pushing down a distinct aggregation. #74810
• Zero row counts could corrupt partition statistics. #74801
• Vector index rewrite could pollute the shared table schema. #74785
• An IllegalStateException during parallel profile collection, fixed by making Tracers fork-aware. #74746
• BE vacuum tasks were not aborted once the FE caller's timeout elapsed. #74694
• Partition consumer errors in ChunksPartitioner were lost instead of being propagated. #74693
• A lock mismatch in blockingAddTabletCtxToScheduler. #74596
• A typo in the azure_adls2_oauth2_client_endpoint configuration field name. #74581
• Pipeline observers were not notified on missed operator state transitions. #74557
• The reported vacuum watermark was incorrect when retain-boundary metadata was gone. #74429
• A data race on MaterializedIndexMeta during updateSchemaBackendId. #74412
• A non-primary-key replica could get stuck with a permanent version hole; it now self-heals. #74408
• A use-after-free of LLVMContext when JIT compilation fails. #74396
• A column mismatch in the missing-replica row of ADMIN SHOW REPLICA STATUS. #74393
• Invalid JIT IR generated for CASE WHEN with mixed float/int WHEN and result types. #74382
• The CatalogRecycleBin was frozen when a cluster snapshot kept failing. #74379
• A partial update targeting a table modified earlier in the same explicit transaction is now rejected with a clear error. #74344
• Immutable-partition updates did not use the transaction's compute resource. #74316
• A potential out-of-bounds error caused by partitioned join. #74315
• Database-level UDFs were not restored to the renamed target database for FE followers. #74313
• Non-root compound predicates yielded EOF instead of NotPushDown. #74218
• Table names were not backquoted when persisting the routine load origStmt. #74188
• An assertion name lookup error in assert-num-rows. #74178
• Aggregation used type-mismatched aggregate functions. #74159
• Force-killed task runs were not archived, and session-prefixed task-run timeouts were not honored. #74146
• RENAME and SWAP (for tables and materialized views) now take the database write lock to avoid concurrent-modification issues. #74100
• Composite-rowset stats were not summed when batching op_writes in primary-key multi-statement transactions. #74059
• The sink was not notified when a distinct aggregate source finished. #74055
• pipe_file_list was not recreated when _statistics_ was dropped. #73970
• A crash in TabletInvertedIndex.deleteTablets, fixed by fast-path skipping empty input. #73955
• The task manager could write an illegal edit log for a task run. #73882
• A set_thread_name race on data directory load threads. #73862
• A race condition in TabletSinkSender::_send_chunk_by_node. #73820
• Incorrect memory accounting in OlapTableSink. #73807
• Incorrect connector bytes-read statistics. #73799
• BACKUP ON (ALL FUNCTION) / (ALL EXTERNAL CATALOGS) failures. #73790
• NullableColumnUnaryFunction lost the decimal scale when all values were null. #73789
• A FlatJSON crash when subwriters had no appends. #73730
• LargeList / FixedSizeList could not be converted to a JSON column during broker load. #73718
• Partial-append of nested types failed during JSON load. #73715
• An NPE in StatisticsCalcUtils when a partition is dropped concurrently. #73711
• Decimal-valued unit counters were not parsed in RuntimeProfileParser. #73683
• An NPE during nested materialized view refresh. #73644
• Multi-statement transactions were not handled correctly in lake publish_log_version. #73423
• An empty ALTER TABLE clause is now rejected, and OPTIMIZE replay was improved. #73352
• Disk cache overflow when Iceberg metadata entries are pinned. #71651

4.0.11

Release Date: June 5, 2026

Behavior Changes

• get_json_string and the other get_json_* functions now return the JSON parse error instead of NULL when implicit VARCHAR-to-JSON parsing fails under ALLOW_THROW_EXCEPTION. The default behavior (returning NULL when the mode is disabled) is unchanged. #73199
• pipeline_enable_large_column_checker is now enabled by default. #72798

Improvements

• Lake write-path load spill files now use a flat, single-level directory layout with the transaction ID baked into each filename, and are reclaimed by a txn-id-based vacuum pass. This moves bulk deletes off the write hot path and lets vacuum clean up spill files leaked by BE crashes. #73064
• SHOW statements (such as SHOW GRANTS and SHOW WAREHOUSES) are now allowed inside an explicit transaction, so BI/JDBC clients that automatically issue SHOW no longer break the transaction flow. #72954
• Java UDAF and UDTF now support STRUCT arguments and return types. #72911
• Scalar Java UDF now supports STRUCT arguments. #72620
• Java UDF now supports DATE and DATETIME types. #72337
• Java UDF now supports nested ARRAY/MAP types. #72283
• Added the FE configuration deploy_serialization_min_thread_pool_size. #72274
• Skipped redundant partition key expression building when an add_partition_value deduplication hit occurs. #73156
• Avoided a redundant latestSnapshot() call in PaimonMetadata#getTableVersionRange. #72892
• Deduplicated commutative AND/OR expressions in scalar operator common subexpression elimination. #72823

Bug Fixes

The following issues have been fixed:

• A memory leak introduced by the UDAF cache. #74025
• An incorrect implementation in aggregate combined functions. #74169
• An issue in shared-data combined txn log mode where the per-partition coordinator claim was not re-recorded on every sender's open, which could drop txn logs. #73962
• A read failure on Iceberg tables that use a custom LocationProvider, fixed by lazily initializing the LocationProvider in SerializableTable. #73482
• A serialization failure caused by the de.javakaffee UnmodifiableCollectionsSerializer, now replaced with a Java 17-compatible version. #73458
• HdfsFsManager copy error messages now include the underlying cause. #73414
• A concurrent SegmentFlushTask race in DeltaWriter::commit(). #73371
• Sort merge provider errors are now propagated to the fragment context instead of being lost. #73337
• An issue where Ranger row-filter/masking policies on Hive views were skipped, so policies on the view or its base tables were not applied. #73265
• Upgraded libthrift to 0.23.0 to address a security vulnerability (CVE). #73243
• An FE file-descriptor leak, fixed by reusing HttpClient instances. #73239
• Parquet broker load errors now include file/column/row context. #73236
• A slot lookup failure for output slots with an empty col_name in the Spark connector external scan. #73225
• A crash in SinkBuffer during graceful exit. #73202
• Query cache conflicts with local shuffle aggregation. #73194
• A use-after-free of the Hive partition descriptor across fragment teardown. #73176
• A thread-safety issue in lake vacuum, fixed by using localtime_r. #73088
• A race condition between PipelineTimerTask doRun and unscheduling during query context destruction. #73082
• Lock contention on read-only query-engine paths, reduced by relaxing DB locks. #73067
• An materialized view refresh failure with SQL Server tables in a JDBC catalog. #72962
• A JNI local-reference leak in JDBCScanner::_init_jdbc_scanner. #72913
• An issue where partition TopN could lose a child's output column. #72848
• An incorrect plan caused by not clearing LambdaArgument.transformedOp before INSERT OVERWRITE re-planning. #72832
• The coordinator lock was held during external resource cleanup. #72830
• Locker rollback is now exception-safe and the unlock order is fixed. #72789
• An incorrect byte order in ColumnDict.merge, now using unsigned byte order. #72778
• A stack-buffer-overflow when formatting into a temporary std::string. #72728
• The HAVING clause is now checked when disabling aggregation spill on a small LIMIT. #72705
• A hang caused by joining forwarded RPCs when draining the runtime_filter worker. #72626
• Incorrect lazy-materialization slot nullability for a materialized view over an outer join. #72621
• merge_condition was not preserved when applying a normal rowset commit. #72542
• Lock contention in TabletScheduler / TabletSchedCtx hot paths during clone, reduced by relaxing DB locks. #72475
• Locker did not roll back a partial intensive-lock acquisition. #72423
• A spillable hash join probe crash. #72397
• COALESCE children are now cast to a common type in the JOIN USING transformer. #72338
• DB READ lock was held too broadly for single-table proc directories, now relaxed to per-table. #72334
• A memory leak when caching the materialized view plan context. #72300
• FSE-v2 did not set the schema for shared-data sorted schema change. #72235
• ConsistencyChecker held a DB READ lock too broadly in periodic scans, now relaxed to per-table READ. #72218
• A BE crash when querying information_schema.warehouse_queries. #72019
• A trailing \r was not stripped before the closing enclose in CRLF CSV inputs. #71866
• Paimon primary key columns were incorrectly marked as non-nullable when querying an external catalog. #71660
• A redundant double slash was created when constructing the JDBC URL if the URI already ended with a trailing slash, breaking strict drivers such as ClickHouse. #70992

4.0.10

Release Date: May 9, 2026

Behavior Changes

• Cloud storage credentials are now redacted in error messages produced by INSERT INTO FILES, preventing accidental exposure of secrets in error logs and SHOW LOAD output. #71245
• StarRocks no longer permits queries against insert-only ACID Hive tables in Hive catalog. Previously such queries could silently return more rows than actually visible because INSERT OVERWRITE operations were not recognized. Affected tables now return an explicit error instead of incorrect results. #71460

Improvements

• Added an Avro schema cache in Iceberg PartitionData construction to remove redundant Jackson ObjectMapper allocations during partition load on tables with many partitions. #72215
• Optimized CatalogRecycleBin.getAdjustedRecycleTimestamp to avoid rebuilding the table-id map on every call, reducing recycle-bin cleanup and tablet scheduling overhead. #72128
• OlapTableSink.createLocation now batches tablet-location lookups in shared-data mode, removing per-tablet StarOS RPCs that previously stalled the planner critical section. #72041
• Java UDAF instances are now loaded and initialized once per query and reused across pipeline driver instances, removing the linear driver-preparation overhead at high pipeline_dop. #72038
• Added BE metrics starrocks_be_staros_shard_info_fallback_total and starrocks_be_staros_shard_info_fallback_failed_total to track when the StarOS worker falls back to fetching shard info from starmgr because the local cache missed. #71620
• File-bundle writes now prefer a tablet-local aggregator so the bundled tablet metadata path does not require cross-node shard-info lookups. #71613
• Audit log entries now include the queried tables and views referenced by each query. #71596
• INSERT INTO FILES CSV export now supports csv.enclose and csv.escape properties for controlling field quoting and escaping. #71589
• Added LDAP direct bind authentication via DN pattern, removing the requirement for an admin search account in single-tenant LDAP setups. #71559
• Added the starrocks_fe_tablet_num metric for shared-data clusters to match the shared-nothing metric set. #71444
• star_mgr_meta_sync_interval_sec is now runtime-mutable via ADMIN SET FRONTEND CONFIG; the new interval takes effect on the next sync cycle without an FE restart. #71675

Bug Fixes

The following issues have been fixed:

• A race in shared-data combined txn log mode where INSERT into per-partition coordinator dispatch could classify legitimate txn logs as orphan and drop them, leaving the transaction stuck in non-VISIBLE state. #72237
• An issue where _incremental_open_node_channel channels in shared-data combined txn log mode silently dropped txn logs because the legacy "sender_id == 0 collects all logs" rule did not apply to incremental channels. #71992
• An issue where RuntimeProfile::to_thrift() could crash BE with std::bad_optional_access when another thread reset counter min/max values during profile serialization. #72904
• An inconsistency in flat JSON merge results when one side contributed empty values. #72973
• An issue where CREATE TABLE for an Iceberg table failed with "Multiple entries with same key: format-version" when the user explicitly specified format-version in PROPERTIES. #72828
• A CompactionScheduler.startCompaction lock scope that held a DB-wide READ lock across single-table critical work, blocking concurrent DDL on other tables in the same database. Switched to IS on DB plus READ on the target table. #72178
• An issue where StarMgrMetaSyncer.syncTableMetaInternal and syncTableColocationInfo held DB READ/WRITE locks across external StarOS RPCs, freezing CREATE/DROP/ALTER/RENAME on every table in the database for the duration of each RPC. #72108
• An issue where StarMgrMetaSyncer.getAllPartitionShardGroupId held the DB READ lock for full iteration over all cloud-native tables and physical partitions, stalling FE threads waiting for the DB write lock on large catalogs. #71614
• A redundant DB READ lock in getTableNamesViewWithLock. The underlying nameToTable is a ConcurrentHashMap, so the enclosing lock added contention without correctness benefit. #72042
• A DB WRITE lock in the read-only /api/{db}/{table}/_count REST endpoint that was unnecessary for computing proximateRowCount(). #72053
• A batch publish deadlock caused by partition version gaps that operations like tablet split, schema change, and alter jobs reserved by advancing nextVersion without a matching publish. #71483
• A deadlock in shared-nothing mode when warming up the LRU cache for rowset metadata while the cache was full. #71459
• A PipelineTimerTask that could remain stuck in waitUtilFinished due to incorrect ordering between consumer registration and finished signaling. #72058
• A condition race in ConnectorSinkPassthroughExchanger::accept that crashed BE with SIGSEGV via out-of-bounds vector access on _writer_count. #71848
• A use-after-free in LoadChannel::get_load_replica_status caused by destruction of a temporary shared_ptr. #71843
• A use-after-free in the information schema sink due to a missing reference count increment in async RPC closure handling. #71513
• A BE crash in reverse(DecimalV3) caused by improper handling of decimal value width. #71834
• A BE crash when UNNEST produced columns whose define-expression carried an ARRAY type, which was incompatible with global dictionary generation downstream. #72027
• An NPE in FE when creating an Iceberg external table with invalid transform argument order such as bucket(4, region); FE now returns a normal analyzer error. #71917
• An issue where Iceberg manifest data file cache entries were missing column statistics when the first query against a table did not request stats (for example SELECT *). #71913
• An issue where the Iceberg min/max optimization was silently skipped when the table was partitioned by bucket(col, N) because PruneHDFSScanColumnRule injected a placeholder materialized column. #71863
• An issue where AggregateJoinPushDownRule failed to rewrite materialized views over Iceberg base tables because Table.getId() was compared instead of identity, and connector-table ids can shift across plan rebuilds. #71856
• An issue where INSERT OVERWRITE into Hive dynamic partitions failed when the metastore listed a partition whose location no longer existed on the file system; the missing partition directory is now created before commit. #71810
• A Parquet scanner failure (Illegal converting from arrow type(dictionary) ...) when Arrow returned dictionary-typed columns, including dictionaries nested inside arrays, structs, and maps. #71855
• An issue where stale scan ranges from earlier batches persisted across ColocatedBackendSelector.Assignment incremental batches, causing files to be re-deployed and re-scanned. #71789
• An issue where PruneShuffleColumnRule did not update the Join outputProperty after pruning Exchange shuffle columns, leading to incorrect downstream distribution. #72003
• Incorrect shuffle distribution caused by a missing project node when PushDownJoinOnExpressionToChildProject was disabled during the first stage of multi-stage materialized view rewrite. #71075
• Duplicate Apply attachments in ReplaceSubqueryRewriteRule when predicate normalization made the same scalar-subquery placeholder appear multiple times. #71155
• A short-circuit issue in EventScheduler where a finished join probe could prevent the pipeline from transitioning to the finished state. #71740
• An issue where AWS assume-role configured via aws.s3.iam_role_arn was not applied to JNI scanners (RCFile / Avro / SequenceFile / Hudi), causing S3 403 errors. #71422
• An issue where Oracle JDBC predicate pushdown produced invalid SQL because date literals did not match the Oracle NLS format; literals are now emitted as date '...'. #71412
• An issue in shared-data mode where a follower FE forwarded DDL to the leader and waited only for FE journal replay, missing the StarMgr journal and producing "no queryable replica" errors for queries that immediately followed table creation. #71263
• An issue where get_tablet_stats for Primary Key tablets repeatedly reloaded the entire TabletMetadata for every segment via get_del_vec_in_meta(). #71672
• An Arrow Flight issue where empty result sets returned column names of r because the placeholder name was emitted instead of the actual schema. #71534
• An issue where parallel_clone_task_per_path updates did not include the store-path count when resizing the CLONE thread pool. #71484
• An issue where the resource group user classifier rejected digit-leading usernames that CREATE USER allowed. The classifier now uses the same validation rule as CREATE USER. #71470
• An issue where HttpServerHandler.channelInactive skipped unregisterConnection when isRegistered() was false, leaking connection-map entries for early-failing requests. #72006
• An issue where Java UDF JNI calls (NewObject, NewArray, NewStringUTF, etc.) did not check for exceptions or null returns, leading to silent failures or undefined behavior. #71734
• An issue where be_tablets.DATA_SIZE reported total_disk_size (including rowset-embedded indexes and the persistent PK index for lake PK tablets) instead of rowset column data bytes. #70735
• A noisy "Failed to batch drop tablets" warning printed by StarMgrMetaSyncer even when there were no shards to delete. #72209
• CVE-2026-42198 (pgjdbc) and CVE-2026-5598 (BouncyCastle): bumped org.postgresql:postgresql to 42.7.11 and BouncyCastle to 1.84. #72797
• CVE in netty: upgraded netty to 4.1.133.Final. #72905
• Cleaned broker CVEs by upgrading netty / jetty / awssdk / jackson dependencies in the broker. #72184
• Upgraded jetty-http to 9.4.58.v20250814 to address known CVEs in the previous jetty-http version. #71762
• Temporarily masked CVE-2026-2332 to unblock the build, since jetty 9.x is EOL and no upstream fix is published. #71914

4.0.9

Release Date: April 16, 2026

Behavior Changes

• When VARBINARY columns appear inside nested types (ARRAY, MAP, or STRUCT), StarRocks now correctly encodes the values in binary format in MySQL result sets. Previously, raw bytes were emitted directly, which could break text-protocol parsing for null bytes or non-printable characters. This change may affect downstream clients or tools that process VARBINARY data inside nested types. #71346
• Routine Load jobs now automatically pause when a non-retryable error is encountered, such as a row causing the Primary Key size limit to be exceeded. Previously, the job would retry indefinitely because such errors were not recognized as non-retryable by the FE transaction status handler. #71161
• SHOW CREATE TABLE and DESC statements now display the Primary Key columns for Paimon external tables. #70535
• Cloud-native tablet metadata fetch operations (such as get_tablet_stats and get_tablet_metadatas) now use a dedicated thread pool instead of the shared UPDATE_TABLET_META_INFO pool. This prevents metadata fetch contention from impacting repair and other tasks. The new thread pool size is configurable via a new BE parameter. #70492

Improvements

• Added session variables to control the encoding behavior of VARBINARY values in MySQL protocol responses, providing fine-grained control over binary result encoding in client connections. #71415
• Added a snapshot_meta.json marker file to cluster snapshots to support integrity validation before snapshot restoration. #71209
• Added warning logs for silently swallowed exceptions in WarehouseManager to improve observability of silent failures. #71215
• Added metrics for Iceberg metadata table queries to support performance monitoring and diagnosis. #70825
• The regexp_replace() function now supports constant folding during FE query planning, reducing planning overhead for queries with constant string arguments. #70804
• Added categorized metrics for Iceberg time travel queries to improve monitoring and performance analysis. #70788
• Added log output when update compaction is suspended, improving visibility into compaction lifecycle. #70538
• SHOW COLUMNS now returns column comments for PostgreSQL external tables. #70520
• Added support for dumping query execution plans when a query encounters an exception, improving diagnosability of runtime failures. #70387
• Tablet deletion during DDL operations is now batched, reducing write lock contention on tablet metadata. #70052
• Added a Force Drop recovery mechanism for synchronous materialized views that are stuck in an error state and cannot be dropped through normal means. #70029

Bug Fixes

The following issues have been fixed:

• An issue where the profile START_TIME and END_TIME were not displayed in the session timezone. #71429
• A shared-object mutation bug in PushDownAggregateRewriter when processing CASE-WHEN/IF expressions, which could cause incorrect query results. #71309
• A use-after-free bug in ThreadPool::do_submit triggered when thread creation fails. #71276
• An issue where information_schema.tables did not properly escape special characters in equality predicates, causing incorrect results. #71273
• An issue where the materialized view scheduler continued to run after the materialized view became inactive. #71265
• Fixed a task signature collision in UpdateTabletSchemaTask across concurrent ALTER jobs that could cause schema update tasks to be skipped. #71242
• An issue where row count estimation produced NaN values for histograms that contained only MCV (Most Common Values) entries. #71241
• A missing dependency on the AWS S3 Transfer Manager in the AWS SDK integration. #71230
• An issue where TaskManager scheduler callbacks did not verify whether the current node is the leader, potentially causing duplicate task execution on follower nodes. #71156
• A thread-local context pollution issue where ConnectContext information was not cleared after a leader-forwarded request completed. #71141
• An issue where the partition predicate was missing in short-circuit point lookups, causing incorrect query results. #71124
• A NullPointerException when analyzing generated columns during Stream Load or Broker Load if a column referenced by the generated column expression was absent from the load schema. #71116
• A use-after-free bug in the error handling path of parallel segment and rowset loading. #71083
• An issue where delvec orphan entries were left behind when a write operation preceded compaction in the same publish batch. #71049
• An issue where queries appeared in the current_queries result via HTTP loopback when checking query progress internally. #71032
• CVE-2026-33870 and CVE-2026-33871. #71017
• A read lock leak in SharedDataStorageVolumeMgr. #70987
• An issue where the input and result columns of the locate() function shared the same NullColumn reference inside BinaryColumns, causing incorrect results. #70957
• An issue where safe tablet deletion checks were incorrectly applied during ALTER operations in share-nothing mode. #70934
• A race condition in _all_global_rf_ready_or_timeout that could prevent global runtime filters from being applied correctly. #70920
• An int32 overflow in the ACCUMULATED metric macro that caused metric values to silently overflow. #70889
• Incorrect aggregation results in dictionary-encoded merge GROUP BY queries. #70866
• CVE-2025-54920. #70862
• A potential data loss issue in aggregation spill caused by incorrect hash table state handling during set_finishing. #70851
• An issue where the content-length header was not reset when proxy_pass_request_body is disabled. #70821
• An issue where the spill directory for load operations was cleaned up in the object destructor rather than during DeltaWriter::close(), potentially causing premature deletion of spill data. #70778
• An issue where INSERT INTO ... BY NAME from FILES() did not correctly push down the schema for partial column sets. #70774
• An issue where connector scan nodes did not reset the scan range source on query retry, causing incorrect results upon retry. #70762
• A potential rowset metadata loss for Primary Key model tablets caused by a GC race during disk re-migration of the form A→B→A. #70727
• An issue where a query-scoped warehouse hint leaked the ComputeResource object in ConnectContext, potentially affecting subsequent queries on the same connection. #70706
• An issue where redundant conjuncts in MySqlScanNode and JDBCScanNode caused BE errors related to VectorizedInPredicate type mismatches. #70694
• A missing libssl-dev dependency in the Ubuntu runtime environment. #70688
• An issue where Iceberg manifest cache completeness was not validated on read, leading to incorrect scan results when the cache was partially populated. #70675
• A duplicate closure reference in _tablet_multi_get_rpc that could cause use-after-free. #70657
• Partial manifest cache writes in the Iceberg ManifestReader that could result in incomplete cache entries and incorrect scan behavior. #70652
• A crash in array_map() when processing arrays that contain null literal elements. #70629
• A stack overflow in the to_base64() function when processing large inputs. #70623
• An issue where INSERT INTO ... BY NAME from FILES() used positional column mapping instead of name-based mapping, causing data to be written to incorrect columns. #70622
• An issue where NOT NULL constraints were incorrectly pushed down into the schema inferred from FILES(), causing load failures for nullable columns. #70621
• An issue where precise external materialized view refresh did not fall back correctly for Iceberg-like connectors. #70589
• A num_short_key_columns mismatch when constructing a partial tablet schema, which could cause data read errors. #70586
• A BE crash that occurred when the child iterator was exhausted in MaskMergeIterator. #70539
• An issue where materialized view refresh jobs repeatedly refreshed partitions whose corresponding Iceberg snapshots had expired. #70523
• An issue where starlet configuration parameters could not be set. #70482
• An issue where the lock-free materialized view rewrite path incorrectly fell back to live metadata, causing inconsistent rewrite behavior. #70475
• An issue in JoinHashTable::merge_ht where dummy rows were not skipped for expression-based join key columns, causing incorrect join results. #70465
• An incorrect equality comparison in InformationFunction that could produce wrong results in certain queries. #70464
• A column type mismatch in the __iceberg_transform_bucket internal function. #70443
• An issue where Iceberg materialized view refresh failed when Iceberg snapshot timestamps were non-monotonic. #70382
• An issue where user authentication credentials were exposed in audit logs and SQL redaction output. #70360
• A CN crash that occurred when scanning an empty tablet with physical split enabled. #70281
• An issue where the VARCHAR column length was not preserved after a redundant CAST was eliminated during query optimization. #70269
• An issue where brpc connection retry logic did not correctly handle a wrapped NoSuchElementException, causing connection failures after the retry attempt. #70203
• An issue where null fractions for outer join columns were not preserved during statistics estimation, leading to suboptimal query plans. #70144
• A memory tracker leak in connector sink operations running on poller threads. #70121

4.0.8

Release Date: March 25, 2026

Behavior Changes

• Improved sql_mode handling: when DIVISION_BY_ZERO or FAIL_PARSE_DATE mode is set, division by zero and date parse failures in str_to_date/str2date now return an error instead of being silently ignored. #70004
• When sql_mode is set to FORBID_INVALID_DATE, invalid dates in INSERT VALUES clauses are now correctly rejected instead of being bypassed. #69803
• Expression partition generated columns are now hidden from DESC and SHOW CREATE TABLE output. #69793
• Client ID is no longer included in audit logs. #69383

Improvements

• Added a configuration item local_exchange_buffer_mem_limit_per_driver to limit the local exchange buffer size to dop * local_exchange_buffer_mem_limit_per_driver. #70393
• Cached file existence check results across versions in check_missing_files to reduce redundant storage I/O. #70364
• Allowed disabling split and reverse scan ranges for descending TopN runtime filters when desc_hint_split_range is set to ≤ 0. #70307
• Added EXPLAIN and EXPLAIN ANALYZE support for INSERT statements in the Trino dialect. #70174
• Optimized Iceberg read performance when position deletes are present. #69717
• Optimized materialized view best-selector strategy based on distributed keys to improve materialized view selection accuracy. #69679

Bug Fixes

The following issues have been fixed:

• JDBC MySQL pushdown failing for unsupported cast operations. #70415
• Type mismatch issues in materialized view refresh. Added mv_refresh_force_partition_type configuration to force partition type in materialized view refresh. #70381
• dataVersion not set correctly when restoring from backup. #70373
• Duplicated partition names in materialized view refresh tasks. #70354
• Incorrect SLF4J parameterized logging using string concatenation instead of placeholder arguments. #70330
• Comment not set when creating Hive tables. #70318
• FileSystemExpirationChecker blocking on slow HDFS close operations. #70311
• Distribute column validation not applied across different partitions in OlapTableSink. #70310
• Constant folding producing INF instead of an error when double addition overflows. #70309
• Typo in Iceberg table creation: field common was used instead of comment. #70267
• Root user not bypassing all Ranger permission checks in some scenarios. #70254
• query_pool memory tracker going negative during data ingestion. #70228
• AuditEventProcessor thread exiting due to OutOfMemoryException. #70206
• SplitTopNRule not applying partition pruning correctly. #70154
• Out-of-bounds access in cal_new_base_version during schema change publish. #70132
• Materialied view rewrite ignoring dropped partitions from the base table. #70130
• Unexpected partition predicate pruning due to type mismatch in boundary comparisons. #70097
• str_to_date losing microsecond precision in BE runtime. #70068
• Join spill process crashing in set_callback_function. #70030
• Broker Load failing GCS authentication after gcs-connector upgrade to version 3.0.13. #70012
• DCHECK failure in DeltaWriter::close() when called from a bthread context. #69960
• Use-after-free race condition in AsyncDeltaWriter close/finish lifecycle. #69940
• Race condition causing write transaction edit log entry to be missed. #69899
• Known CVE vulnerabilities. #69863
• Follower FE not waiting for journal replay in changeCatalogDb. #69834
• Incorrect LIKE pattern matching with backslash escape sequences. #69775
• Expression analysis failure after renaming a partition column. #69771
• Use-after-free crash in AsyncDeltaWriter::close. #69770
• Crash in local partition TopN execution. #69752
• Incorrect behavior in PartitionColumnMinMaxRewriteRule caused by Partition.hasStorageData. #69751
• Duplicated CSV compression suffix in file sink output filenames. #69749
• lake_capture_tablet_and_rowsets not gated behind an experimental configuration flag. #69748
• Incorrect partition min pruning with shadow partitions. #69641
• Java UDTF/UDAF crashing when method parameters use generic types. #69197
• Per-query metadata not released after query planning, causing FE OOM during concurrent query execution. #68444
• Query-scope warehouse hint leaking ComputeResource in ConnectContext. #70706
• Lock-free materialized view rewrite incorrectly falling back to live metadata. #70475
• Duplicate closure reference in _tablet_multi_get_rpc. #70657
• Infinite recursion in ReplaceColumnRefRewriter. #66974
• NOT NULL constraint incorrectly pushed down to FILES() table function schema. #70621
• num_short_key_columns mismatch in partial tablet schema. #70586
• COLUMN_UPSERT_MODE checksum error in shared-data clusters. #65320
• Column type mismatch for __iceberg_transform_bucket. #70443
• Starlet configuration items not taking effect. #70482
• DCG data not read correctly when switching from column mode to row mode in partial update. #61529

4.0.7

Release Date: March 12, 2026

Behavior Change

• Disallowed creating materialized views based on Iceberg views. #69471
• Fixed inconsistencies in multi-statement Stream Load transaction behavior. #68542

Improvements

• Added fine-grained trace counters for LakePersistentIndex in the Publish phase. #69640
• Triggered early flush in LakePersistentIndex when rebuild row count exceeds the threshold. #69698
• Added dump_lake_persistent_index_sst operation to meta_tool. #69682
• Improved REPAIR TABLE functionality and SHOW TABLET status display. #69656
• Supports ADMIN SHOW TABLET STATUS for cloud-native tables. #69616
• Upgraded hadoop-client from 3.4.2 to 3.4.3. #69503
• Prevented crashes when deserialization mismatches occur. #69481
• Pushed down predicates to FE when querying information_schema.loads. #69472
• Optimized the SQL displayed for materialized view refresh TaskRuns. #69437
• Bypassed caching in CachingIcebergCatalog when vended credentials are enabled. #69434
• Uses tryLock with timeout in canTxnFinished to reduce lock contention. #69427
• Added a global readonly variable @@run_mode. #69247
• Uses Estimator to estimate cache entry weight in DeltaLakeMetastore. #69244
• Added resource share type support for AWS Glue GetDatabases API. #69056
• Extracted range predicates from scalar-subqueries containing convert_tz. #69055
• Added ByteBuffer Estimator. #69042
• Supports fast cancel for Lake DeltaWriter in shared-data clusters. #68877
• Supports an interface to add physical partitions for random distribution tables. #68503
• Gated SQL transactions behind the session variable enable_sql_transaction (default: true). #63535
• Added partition scan number limit when querying external tables. #68480

Bug Fixes

The following issues have been fixed:

• Incorrect value for the metric g_publish_version_failed_tasks when the resource is busy. #69526
• NPE in IcebergCatalog.getPartitionLastUpdatedTime when the snapshot has expired. #68925
• DCHECK failure in DeltaWriter::close() when called from bthread context. #70057
• Several use-after-free issues. #69968
• Use-after-free race in AsyncDeltaWriter close/finish lifecycle. #69961
• Corrupted cache for PK SST tables is not clraered. #69693
• AsyncFlushOutputStream use-after-free issue. #69688
• Retention clock reset issue and incomplete scan in disableRecoverPartitionWithSameName. #69677
• NPE in StreamLoadMultiStmtTask.cancelAfterRestart after deserialization. #69662
• Unnecessary RPCs and metadata queries caused by the incorrect logic of SchemaBeTabletsScanner. #69645
• Graceful exit caused different transactions to publish the same version. #69639
• TabletUpdates::get_column_values crashes with SIGSEGV when the Primary Key Index contains stale entries that point to rowsets that have been compacted. #69617
• KILL ANALYZE fails to stop ANALYZE TABLE tasks. #69592
• Unexpected behavior because not all exceptions of RowGroupWriter are caught. #69568
• TaskRun warehouse display issues after changing the warehouse for the materialized view. #69567
• Sort key does not include newly added key columns after schema change on aggregate and unique tables. #69529
• Issues caused by isInternalCancelError using equals. #69523
• TaskManager scheduling bugs after ALTER MATERIALIZED VIEW. #69504
• Pipeline will be blocked or crash because not all exceptions of ParquetFileWriter::close are caught. #69492
• Materialized view force refresh bugs for partitioned tables. #69488
• Incorrect status was returned when certain writers failed to flush data. #69473
• Rowset files were deleted when Primary Key tablets were moved to trash. #69438
• INSERT failure when the range of an automatic partition is enclosed by an existing merged partition. #69429
• Materialized view tablet meta inconsistency between FE leader and follower. #69428
• Concurrency bugs related to function fields. #69315
• Premature deletion of data because rollup handler's active transaction ID is not considered in computeMinActiveTxnId. #69285
• Lock leak in addPartitions caused by name-based table lookup after concurrent SWAP. #69284
• DROP FUNCTION IF EXISTS ignored the ifExists flag. #69216
• Inconsistent behavior between StarRocks and MySQL-compatible syntax when CAST(... AS SIGNED) in TypeParser. #69181
• Issue with case-insensitive partition lookup in query table copy. #69173
• Missing aggregate function when MIN/MAX stats rewrite failed. #69149
• CVE-2025-67721. #69138
• All-null value handling bug in synchronous materialized views. #69136
• Projection loss in materialized view rewrite due to shared mutable state. #69063
• Incorrect estimation of the Iceberg cache Weigher. #69058
• FULL OUTER JOIN USING issue with constant subqueries. #69028
• Issue with DISTINCT ORDER BY alias resolution for duplicated constants. #69014
• Issue when materialized view visits external catalog on reload. #68926
• NPE in Iceberg getPartitions. #68907
• The container were not properly included in the Azure ABFS/WASB FileSystem cache key. #68901
• Erroneous query results after modifying CHAR column length in shared-data clusters. #68808
• Case-insensitive issue with username in LDAP authentication. #67966
• Partitions could not be created after adding storage_cooldown_ttl to a table. #60290

4.0.6

Release Date: February 14, 2026

Improvements

• Support Partition Transforms with parentheses when creating Iceberg tables (for example, PARTITION BY (bucket(k1, 3))). #68945
• Removed the restriction that partition columns in Iceberg tables must be at the end of the column list; they can now be defined at any position. #68340
• Introduced host-level sorting for Iceberg table sink, controlled by the system variable connector_sink_sort_scope (Default: FILE), to organize data layout for better read performance. #68121
• Improved error messages for Iceberg partition transform functions (for example, bucket, truncate) when the argument count is incorrect. #68349
• Refactored table property handling to improve support for different file formats (ORC/Parquet) and compression codecs in Iceberg tables. #68588
• Added table-level query timeout configuration table_query_timeout for fine-grained control (Priority: Session > Table > Cluster). #67547
• Supports the ADMIN SHOW AUTOMATED CLUSTER SNAPSHOT statement to view automated snapshot status and schedule. #68455
• Supports displaying the original user-defined SQL with comments in SHOW CREATE VIEW. #68040
• Exposed Merge Commit-enabled Stream Load tasks in information_schema.loads for better observability. #67879
• Introduced FE memory estimation utility API /api/memory_usage. #68287
• Reduced unnecessary logging in CatalogRecycleBin during partition recycling. #68533
• Triggered refresh of related asynchronous materialized views when the base table undergoes Swap/Drop/Replace Partition operations. #68430
• Supports VARBINARY type for count distinct-like aggregate functions. #68442
• Enhanced expression statistics to propagate histogram MCV for semantics-safe expressions (for example, cast(k as bigint) + 10) to improve skew detection. #68292

Bug Fixes

The following issues have been fixed:

• Potential crashes in Skew Join V2 runtime filters. #67611
• Join predicate type mismatch (for example, INT = VARCHAR) caused by low-cardinality rewriting. #68568
• Issues in query queue allocation time and pending timeout logic. #65802
• unique_id conflict for Flat JSON extended columns after schema changes. #68279
• Concurrent partition access issues in OlapTableSink.complete(). #68853
• Incorrect metadata tracking when restoring manually downloaded cluster snapshots. #68368
• Double slashes in backup paths when the repository location ends with /. #68764
• OBS AK/SK credentials in the SHOW CREATE CATALOG output were not masked. #65462

4.0.5

Release Date: February 3, 2026

Improvements

• Bumped Paimon version to 1.3.1. #67098
• Restored missing optimizations in DP statistics estimation to reduce redundant calculations. #67852
• Improved pruning in DP Join reorder to skip expensive candidate plans earlier. #67828
• Optimized JoinReorderDP partition enumeration to reduce object allocation and added an atom count cap (≤ 62). #67643
• Optimized DP join reorder pruning and added checks to BitSet to reduce stream operation overhead. #67644
• Skipped predicate column statistics collection during DP statistics estimation to reduce CPU overhead. #67663
• Optimized correlated Join row count estimation to avoid repeatedly building Statistics objects. #67773
• Reduced memory allocations in Statistics.getUsedColumns. #67786
• Avoided redundant Statistics map copies when only row counts are updated. #67777
• Skipped aggregate pushdown logic when no aggregation exists in the query to reduce overhead. #67603
• Improved COUNT DISTINCT over windows, added support for fused multi-distinct aggregations, and optimized CTE generation. #67453
• Supports map_agg function in the Trino dialect. #66673
• Supports batching retrieval of LakeTablet location information during physical planning to reduce RPC calls in shared-data clusters. #67325
• Added a thread pool for Publish Version transactions to shared-nothing clusters to improve concurrency. #67797
• Optimized LocalMetastore locking granularity by replacing database-level locks with table-level locks. #67658
• Refactored MergeCommitTask lifecycle management and added support for task cancellation. #67425
• Supports intervals for automated cluster snapshots. #67525
• Automatically cleaned up unused mem_pool entries in MemTrackerManager. #67347
• Ignored information_schema queries during warehouse idle checks. #67958
• Supports dynamically enabling global shuffle for Iceberg table sinks based on data distribution. #67442
• Added Profile metrics for connector sink modules. #67761
• Improved the collection and display of load spill metrics in Profiles, distinguishing between local and remote I/O. #67527
• Changed Async-Profiler log level to Error to avoid repeating warning logs. #67297
• Notified Starlet during BE shutdown to report SHUTDOWN status to StarMgr. #67461

Bug Fixes

The following issues have been fixed:

• Lacking support for legal simple paths containing hyphens (-). #67988
• Runtime error when aggregate pushdown occurred on grouping keys involving JSON types. #68142
• Issue where JSON path rewrite rules incorrectly pruned partition columns referenced in partition predicates. #67986
• Type mismatch issue when rewriting simple aggregation using statistics. #67829
• Potential heap-buffer-overflow in partition Joins. #67435
• Duplicate slot_ids introduced when pushing down heavy expressions. #67477
• Division-by-zero error in ExecutionDAG fragment connection for lacking precondition checks. #67918
• Potential issues caused by fragment parallel prepare for single BE. #67798
• Operator terminates incorrectly for lacking set_finished method for RawValuesSourceOperator. #67609
• BE crash caused by unsupported DECIMAL256 type (precision > 38) in column aggregators. #68134
• Shared-data clusters lack support for Fast Schema Evolution v2 over DELETE operations by carrying schema_key in requests. #67456
• Shared-data clusters lack support for Fast Schema Evolution v2 over synchronous materialized views and traditional schema changes. #67443
• Vacuum might accidentally delete files when file bundling is disabled during FE downgrade. #67849
• Incorrect graceful exit handling in MySQLReadListener. #67917

4.0.4

Release Date: January 16, 2026

Improvements

• Supports Parallel Prepare for Operators and Drivers, and single-node batch fragment deployment to improve query scheduling performance. #63956
• Optimized deltaRows calculation with lazy evaluation for large partition tables. #66381
• Optimized Flat JSON processing with sequential iteration and improved path derivation. #66941 #66850
• Supports releasing Spill Operator memory earlier to reduce memory usage in group execution. #66669
• Optimized the logic to reduce string comparison overhead. #66570
• Improved skew detection in GroupByCountDistinctDataSkewEliminateRule and SkewJoinOptimizeRule to support histogram and NULL-based strategies. #66640 #67100
• Enhanced Column ownership management in Chunk using Move semantics to reduce Copy-On-Write overhead. #66805
• For shared-data clusters, added FE TableSchemaService and updated MetaScanNode to support Fast Schema Evolution v2 schema retrieval. #66142 #66970
• Supports multi-warehouse Backend resource statistics and parallelism (DOP) calculation for better resource isolation. #66632
• Supports configuring Iceberg split size via StarRocks session variable connector_huge_file_size. #67044
• Supports label-formatted statistics in QueryDumpDeserializer. #66656
• Added an FE configuration lake_enable_fullvacuum (Default: false) to allow disabling Full Vacuum in shared-data clusters. #63859
• Upgraded lz4 dependency to v1.10.0. #67045
• Added fallback logic for sample-type cardinality estimation when row count is 0. #65599
• Validated Strict Weak Ordering property for lambda comparator in array_sort. #66951
• Optimized error messages when fetching external table metadata (Delta/Hive/Hudi/Iceberg) fails, showing root causes. #66916
• Supports dumping pipeline status on query timeout and cancelling with TIMEOUT state in FE. #66540
• Displays matched rule index in SQL blacklist error messages. #66618
• Added labels to column statistics in EXPLAIN output. #65899
• Filtered out "cancel fragment" logs for normal query completions (for example, LIMIT reached). #66506
• Reduced Backend heartbeat failure logs when the warehouse is suspended. #66733
• Supports IF EXISTS in the ALTER STORAGE VOLUME syntax. #66691

Bug Fix

The following issues have been fixed:

• Incorrect DISTINCT and GROUP BY results under Low Cardinality optimization due to missing withLocalShuffle. #66768
• Rewrite error for JSON v2 functions with Lambda expressions. #66550
• Incorrect application of Partition Join in Null-aware Left Anti Join within correlated subqueries. #67038
• Incorrect row count calculation in the Meta Scan rewrite rule. #66852
• Nullable property mismatched in Union Node when rewriting Meta Scan by statistics. #67051
• BE crash caused by optimization logic for Ranking window functions when PARTITION BY and ORDER BY are missing. #67094
• Potential wrong results in Group Execution Join with window functions. #66441
• Incorrect results from PartitionColumnMinMaxRewriteRule under specific filter conditions. #66356
• Incorrect Nullable property deduction in Union operations after aggregation. #65429
• Crash in percentile_approx_weighted when handling compression parameters. #64838
• Crash when spilling with large string encoding. #61495
• Crash triggered by multiple calls to set_collector when pushing down local TopN. #66199
• Dependency deduction error in LowCardinality rewrite logic. #66795
• Rowset ID leak when rowset commit fails. #66301
• Metacache lock contention. #66637
• Ingestion failure when column-mode partial update is used with conditional update. #66139
• Concurrent import failure caused by Tablet deletion during the ALTER operation. #65396
• Tablet metadata load error due to RocksDB iteration timeout. #65146
• Compression settings were not applied during table creation and Schema Change in shared-data clusters. #65673
• Delete Vector CRC32 compatibility issue during upgrade. #65442
• Status check logic error in file cleanup after clone task failure. #65709
• Abnormal statistics collection logic after INSERT OVERWRITE. #65327 #65298 #65225
• Foreign Key constraints were lost after FE restart. #66474
• Metadata retrieval error after Warehouse deletion. #66436
• Inaccurate Audit Log scan statistics under high selectivity filters. #66280
• Incorrect query error rate metrics calculation logic. #65891
• Potential MySQL connection leaks when tasks exit. #66829
• BE status was not updated immediately on the SIGSEGV crash. #66212
• NPE during LDAP user login. #65843
• Inaccurate error log when switching users in HTTP SQL requests. #65371
• HTTP context leaks during TCP connection reuse. #65203
• Missing QueryDetail in Profile logs for queries forwarded from Follower. #64395
• Missing Prepare/Execute details in Audit logs. #65448
• Crash caused by HyperLogLog memory allocation failures. #66747
• Issue with the trim function memory reservation. #66477 #66428
• CVE-2025-66566 and CVE-2025-12183. #66453 #66362 #67053
• Race condition in Exec Group driver submission. #66099
• Use-after-free risk in Pipeline countdown. #65940
• MemoryScratchSinkOperator hangs when the queue closes. #66041
• Filesystem cache key collision issue. #65823
• Wrong subtask count in SHOW PROC '/compactions'. #67209
• A unified JSON format is not returned in the Query Profile API. #67077
• Improper getTable exception handling that affects the materialized view check. #67224
• Inconsistent output of the Extra column from the DESC statement for native and cloud-native tables. #67238
• Race condition in single-node deployments. #67215
• Log leakage from third-party libraries. #67129
• Incorrect REST Catalog authentication logic that causes authentication failures. #66861

4.0.3

Release Date: December 25, 2025

Improvements

• Supports ORDER BY clauses for STRUCT data types #66035
• Supports creating Iceberg views with properties and displaying properties in the output of SHOW CREATE VIEW. #65938
• Supports altering Iceberg table partition specs using ALTER TABLE ADD/DROP PARTITION COLUMN. #65922
• Supports COUNT/SUM/AVG(DISTINCT) aggregation over framed windows (for example, ORDER BY/PARTITION BY) with optimization options. #65815
• Optimized CSV parsing performance by using memchr for single-character delimiters. #63715
• Added an optimizer rule to push down Partial TopN to the Pre-Aggregation phase to reduce network overhead. #61497
• Enhanced Data Cache monitoring
  • Added new metrics for memory/disk quota and usage. #66168
  • Added Page Cache statistics to the api/datacache/stat HTTP endpoint. #66240
  • Added hit rate statistics for native tables. #66198
• Optimized Sort and Aggregation operators to support rapid memory release in OOM scenarios. #66157
• Added TableSchemaService in FE for shared-data clusters to allow CNs to fetch specific schemas on demand. #66142
• Optimized Fast Schema Evolution to retain history schemas until all dependent ingestion jobs are finished. #65799
• Enhanced filterPartitionsByTTL to properly handle NULL partition values to prevent all partitions from being filtered. #65923
• Optimized FusedMultiDistinctState to clear the associated MemPool upon reset. #66073
• Made ICEBERG_CATALOG_SECURITY property check case-insensitive in Iceberg REST Catalog. #66028
• Added HTTP endpoint GET /service_id to retrieve StarOS Service ID in shared-data clusters. #65816
• Replaced deprecated metadata.broker.list with bootstrap.servers in Kafka consumer configurations. #65437
• Added FE configuration lake_enable_fullvacuum (Default: false) to allow disabling the Full Vacuum Daemon. #66685
• Updated lz4 library to v1.10.0. #67080

Bug Fixes

The following issues have been fixed:

• latest_cached_tablet_metadata could cause versions to be incorrectly skipped during batch Publish. #66558
• Potential issues caused by ClusterSnapshot relative checks in CatalogRecycleBin when running in shared-nothing clusters. #66501
• BE crash when writing complex data types (ARRAY/MAP/STRUCT) to Iceberg tables during Spill operations. #66209
• Potential hang in Connector Chunk Sink when the writer's initialization or initial write fails. #65951
• Connector Chunk Sink bug where PartitionChunkWriter initialization failure caused a null pointer dereference during close. #66097
• Setting a non-existent system variable would silently succeed instead of reporting an error. #66022
• Bundle metadata parsing failure when Data Cache is corrupted. #66021
• MetaScan returned NULL instead of 0 for count columns when the result is empty. #66010
• SHOW VERBOSE RESOURCE GROUP ALL displays NULL instead of default_mem_pool for resource groups created in earlier versions. #65982
• A RuntimeException during query execution after disabling the flat_json table configuration. #65921
• Type mismatch issue in shared-data clusters caused by rewriting min/max statistics to MetaScan after Schema Change. #65911
• BE crash caused by ranking window optimization when PARTITION BY and ORDER BY are missing. #67093
• Incorrect can_use_bf check when merging runtime filters, which could lead to wrong results or crashes. #67062
• Pushing down runtime bitset filters into nested OR predicates causes incorrect results. #67061
• Potential data race and data loss issues caused by write or flush operations after the DeltaWriter has finished. #66966
• Execution error caused by mismatched nullable properties when rewriting simple aggregation to MetaScan. #67068
• Incorrect row count calculation in the MetaScan rewrite rule. #66967
• Versions might be incorrectly skipped during batch Publish due to inconsistent cached tablet metadata. #66575
• Improper error handling for memory allocation failures in HyperLogLog operations. #66827

4.0.2

Release Date: December 4, 2025

New Features

• Introduced a new resource group attribute, mem_pool, allowing multiple resource groups to share the same memory pool and enforce a joint memory limit for the pool. This feature is backward compatible. default_mem_pool is used if mem_pool is not specified. #64112

Improvements

• Reduced remote storage access during Vacuum after File Bundling is enabled. #65793
• The File Bundling feature caches the latest tablet metadata. #65640
• Improved safety and stability for long-string scenarios. #65433 #65148
• Optimized the SplitTopNAggregateRule logic to avoid performance regression. #65478
• Applied the Iceberg/DeltaLake table statistics collection strategy to other external data sources to avoid collecting statistics when the table is a single table. #65430
• Added Page Cache metrics to the Data Cache HTTP API api/datacache/app_stat. #65341
• Supports ORC file splitting to enable parallel scanning of a single large ORC file. #65188
• Added selectivity estimation for IF predicates in the optimizer. #64962
• Supports constant evaluation of hour, minute, and second for DATE and DATETIME types in the FE. #64953
• Enabled rewrite of simple aggregation to MetaScan by default. #64698
• Improved multiple-replica assignment handling in shared-data clusters for enhanced reliability. #64245
• Exposes cache hit ratio in audit logs and metrics. #63964
• Estimates per-bucket distinct counts for histograms using HyperLogLog or sampling to provide more accurate NDV for predicates and joins. #58516
• Supports FULL OUTER JOIN USING with SQL-standard semantics. #65122
• Prints memory information when Optimizer times out for diagnostics. #65206

Bug Fixes

The following issues have been fixed:

• DECIMAL56 mod-related issue. #65795
• Issue related to Iceberg scan range handling. #65658
• MetaScan rewrite issues on temporary partitions and random buckets. #65617
• JsonPathRewriteRule uses the wrong table after transparent materialized view rewrite. #65597
• Materialized view refresh failures when partition_retention_condition referenced generated columns. #65575
• Iceberg min/max value typing issue. #65551
• Issue with queries against information_schema.tables and views across different databases when enable_evaluate_schema_scan_rule is set to true. #65533
• Integer overflow in JSON array comparison. #64981
• MySQL Reader does not support SSL. #65291
• ARM build issue caused by SVE build incompatibility. #65268
• Queries based on bucket-aware execution may get stuck for bucketed Iceberg tables. #65261
• Robust error propagation and memory safety issues for the lack of memory limit checks in OLAP table scan. #65131

Behavior Changes

• When a materialized view is inactivated, the system recursively inactivates its dependent materialized views. #65317
• Uses the original materialized view query SQL (including comments/formatting) when generating SHOW CREATE output. #64318

4.0.1

Release Date: November 17, 2025

Improvements

• Optimized TaskRun session variable handling to process known variables only. #64150
• Supports collecting statistics of Iceberg and Delta Lake tables from metadata by default. #64140
• Supports collecting statistics of Iceberg tables with bucket and truncate partition transform. #64122
• Supports inspecting FE /proc profile for debugging. #63954
• Enhanced OAuth2 and JWT authentication support for Iceberg REST catalogs. #63882
• Improved bundle tablet metadata validation and recovery handling. #63949
• Improved scan-range memory estimation logic. #64158

Bug Fixes

The following issues have been fixed:

• Transaction logs were deleted when publishing bundle tablets. #64030
• The join algorithm cannot guarantee the sort property because, after joining, the sort property is not reset. #64086
• Issues related to transparent materialized view rewrite. #63962

Behavior Changes

• Added the property enable_iceberg_table_cache to Iceberg Catalogs to optionally disable Iceberg table cache and allow it always to read the latest data. #64082
• Ensured INSERT ... SELECT reads the freshest metadata by refreshing external tables before planning. #64026
• Increased lock table slots to 256 and added rid to slow-lock logs. #63945
• Temporarily disabled shared_scan due to incompatibility with event-based scheduling. #63543
• Changed the default Hive Catalog cache TTL to 24 hours and removed unused parameters. #63459
• Automatically determine the Partial Update mode based on the session variable and the number of inserted columns. #62091

4.0.0

Release date: October 17, 2025

Data Lake Analytics

• Unified Page Cache and Data Cache for BE metadata, and adopted an adaptive strategy for scaling. #61640
• Optimized metadata file parsing for Iceberg statistics to avoid repetitive parsing. #59955
• Optimized COUNT/MIN/MAX queries against Iceberg metadata by efficiently skipping over data file scans, significantly improving aggregation query performance on large partitioned tables and reducing resource consumption. #60385
• Supports compaction for Iceberg tables via procedure rewrite_data_files.
• Supports Iceberg tables with hidden partitions, including creating, writing, and reading the tables. #58914
• Supports setting sort keys when creating Iceberg tables.
• Optimizes sink performance for Iceberg tables.
  • Iceberg Sink supports spilling large operators, global shuffle, and local sorting to optimize memory usage and address small file issues. #61963
  • Iceberg Sink optimizes local sorting based on Spill Partition Writer to improve write efficiency. #62096
  • Iceberg Sink supports global shuffle for partitions to further reduce small files. #62123
• Enhanced bucket-aware execution for Iceberg tables to improve concurrency and distribution capabilities of bucketed tables. #61756
• Supports the TIME data type in the Paimon catalog. #58292
• Upgraded Iceberg version to 1.10.0. #63667

Security and Authentication

• In scenarios where JWT authentication and the Iceberg REST Catalog are used, StarRocks supports the passthrough of user login information to Iceberg via the REST Session Catalog for subsequent data access authentication. #59611 #58850
• Supports vended credentials for the Iceberg catalog.
• Supports granting StarRocks internal roles to external groups obtained via Group Provider. #63385 #63258
• Added REFRESH privilege to external tables to control the permission to refresh them. #63385

Storage Optimization and Cluster Management

• Introduced  the File Bundling optimization for the cloud-native table in shared-data clusters to automatically bundle the data files generated by loading, Compaction, or Publish operations, thereby reducing the API cost caused by high-frequency access to the external storage system. File Bundling is enabled by default for tables created in v4.0 or later. #58316
• Supports Multi-Table Write-Write Transaction to allow users to control the atomic submission of INSERT, UPDATE, and DELETE operations. The transaction supports Stream Load and INSERT INTO interfaces, effectively guaranteeing cross-table consistency in ETL and real-time write scenarios. #61362
• Supports Kafka 4.0 for Routine Load.
• Supports full-text inverted indexes on Primary Key tables in shared-nothing clusters.
• Supports modifying aggregate keys of Aggregate tables. #62253
• Supports enabling case-insensitive processing on names of catalogs, databases, tables, views, and materialized views. #61136
• Supports blacklisting Compute Nodes in shared-data clusters. #60830
• Supports global connection ID. #57256
• Added the recyclebin_catalogs metadata view to Information Schema to display recoverable deleted metadata. #51007

Query and Performance Improvement

• Supports DECIMAL256 data type, expanding the upper limit of precision from 38 to 76 bits. Its 256-bit storage provides better adaptability to high-precision financial and scientific computing scenarios, effectively mitigating DECIMAL128's precision overflow problem in very large aggregations and high-order operations. #59645
• Improved the performance for basic operators.#61691 #61632 #62585 #61405 #61429
• Optimized the performance of the JOIN and AGG operators. #61691
• [Preview] Introduced SQL Plan Manager to allow users to bind a query plan to a query, thereby preventing the query plan from changing due to system state changes (mainly data updates and statistics updates), thus stabilizing query performance. #56310
• Introduced Partition-wise Spillable Aggregate/Distinct operators to replace the original Spill implementation based on sorted aggregation, significantly improving aggregation performance and reducing read/write overhead in complex and high-cardinality GROUP BY scenarios. #60216
• Flat JSON V2:
  • Supports configuring Flat JSON on the table level. #57379
  • Enhance JSON columnar storage by retaining the V1 mechanism while adding page- and segment-level indexes (ZoneMaps, Bloom filters), predicate pushdown with late materialization, dictionary encoding, and integration of a low-cardinality global dictionary to significantly boost execution efficiency. #60953
• Supports an adaptive ZoneMap index creation strategy for the STRING data type. #61960
• Enhanced query observability:
  • Optimized EXPLAIN ANALYZE output to display the execution metrics by group and by operator for better readability. #63326
  • QueryDetailActionV2 and QueryProfileActionV2 now support JSON format, enhancing cross-FE query capabilities. #63235
  • Supports retrieving Query Profile information across all FEs. #61345
  • SHOW PROCESSLIST statements display Catalog, Query ID, and other information. #62552
  • Enhanced query queue and process monitoring, supporting display of Running/Pending statuses.#62261
• Materialized view rewrites consider the distribution and sort keys of the original table, improving the selection of optimal materialized views. #62830

Functions and SQL Syntax

• Added the following functions:
  • bitmap_hash64 #56913
  • bool_or #57414
  • strpos #57278
  • to_datetime and to_datetime_ntz #60637
  • regexp_count #57182
  • tokenize #58965
  • format_bytes #61535
  • encode_sort_key #61781
  • column_size and column_compressed_size #62481
• Provides the following syntactic extensions:
  • Supports IF NOT EXISTS keywords in CREATE ANALYZE FULL TABLE. #59789
  • Supports EXCLUDE clauses in SELECT. #57411
  • Supports FILTER clauses in aggregate functions, improving readability and execution efficiency of conditional aggregations. #58937

Behavior Changes

• Adjust the logic of the materialized view parameter auto_partition_refresh_number to limit the number of partitions to refresh regardless of auto refresh or manual refresh. #62301
• Flat JSON is enabled by default. #62097
• The default value of the system variable enable_materialized_view_agg_pushdown_rewrite is set to true, indicating that aggregation pushdown for materialized view query rewrite is enabled by default. #60976
• Changed the type of some columns in information_schema.materialized_views to better align with the corresponding data. #60054
• The split_part function returns NULL when the delimiter is not matched. #56967
• Use STRING to replace fixed-length CHAR in CTAS/CREATE MATERIALIZED VIEW to avoid deducing the wrong column length, which may cause materialized view refresh failures. #63114 #62476
• Data Cache-related configurations are simplified. #61640
  • datacache_mem_size and datacache_disk_size are now effective.
  • storage_page_cache_limit, block_cache_mem_size, block_cache_disk_size are deprecated.
• Added new catalog properties (remote_file_cache_memory_ratio for Hive, and iceberg_data_file_cache_memory_usage_ratio and iceberg_delete_file_cache_memory_usage_ratio for Iceberg) to limit the memory resources used for Hive and Iceberg metadata cache, and set the default values to 0.1 (10%). Adjust the metadata cache TTL to 24 hours. #63459 #63373 #61966 #62288
• SHOW DATA DISTRIBUTION now will not merge the statistics of all materialized indexes with the same bucket sequence number. It only shows data distribution at the materialized index level. #59656
• The default bucket size for automatic bucket tables is changed from 4GB to 1GB to improve performance and resource utilization. #63168
• The system determines the Partial Update mode based on the corresponding session variable and the number of columns in the INSERT statement. #62091
• Optimized the fe_tablet_schedules view in the Information Schema. #62073 #59813
  • Renamed the TABLET_STATUS column to SCHEDULE_REASON, the CLONE_SRC column to SRC_BE_ID, and the CLONE_DEST column to DEST_BE_ID.
  • The data types of the CREATE_TIME, SCHEDULE_TIME and FINISH_TIME columns have been changed from DOUBLE to DATETIME.
• The is_leader label has been added to some FE metrics. #63004
• Shared-data clusters using Microsoft Azure Blob Storage and Data Lake Storage Gen 2 as object storage will experience Data Cache failure after being upgraded to v4.0. The system will automatically reload the cache.