All Products
Search

This Product

    E-MapReduce:Kernel release notes

    Document Center

    E-MapReduce:Kernel release notes

    Last Updated:Nov 03, 2025

    This topic describes the feature updates of the EMR Serverless StarRocks kernel.

    3.3

    Note

    EMR Serverless StarRocks 3.3 is based on the Linux Foundation Project StarRocks. For more information, see StarRocks version 3.3.

    Version 3.3.13-1.1.1

    Release date

    September 21, 2025

    Updates

    Feature category

    Description

    Base engine

    [Optimized features]

    • Optimized the performance of collecting partition statistics.

    • Support added for dynamic adjustment of the thread pool size for the frontend (FE) query fragments deployer.

    • Supports configurable FE checkpoint parameters and logs long-running tasks.

    [Fixed issues]

    • The number of Java Database Connectivity (JDBC) connections would exceed the jdbc_connection_pool_size limit.

    • Dynamic partition creation would fail due to partition name conflicts.

    • Scan data information was missing from the statistics of import tasks.

    • Inverted index queries incorrectly filtered null fields.

    • Full-text search would return an incorrect result if the pattern for a regular expression match was an empty string.

    • Insert overwrite operations would fail on manually created partitions.

    • An Alter Job was canceled due to a conflict with partition creation.

    • The status of an analyze job for statistics collection was abnormal.

    • Import transaction logs on the FE were incomplete.

    Storage and compute separation

    [Fixed issues]

    • [Important] A graceful exit of a backend (BE) or compute node (CN) would make tablet metadata files unreadable.

    • The max compaction score metric on the leader FE was not updated.

    • Transaction logs would be lost during concurrent batch publish and schema change operations.

    Lakehouse

    [Fixed issues]

    • The UUID format in Paimon tables was processed incorrectly.

    • User authentication was incorrect for Paimon materialized views (MVs) in Data Lake Formation (DLF) 2.5.

    Version 3.3.13-1.1.0

    Release date

    July 21, 2025

    Updates

    Feature category

    Description

    Base engine

    [Optimized features]

    • Adds fe.load logs to improve the observability of the data loading process.

    • Adds the sink_bytes field to the information_schema.loads table.

    • Optimizes Routine Load error logs by adding key information such as offsets to facilitate troubleshooting.

    [Fixed issues]

    • Fixes an issue where global_state_manager in ConnectContext might be null. This prevents query failures, import failures, or missing profiles.

    • Fixes an issue of missing column statistics to ensure data accuracy.

    • Fixes the Query version not found error.

    • Adds the slow_lock_print_stack parameter to prevent process stuttering caused by thread stack printing in large-scale clusters.

    • Fixes an out-of-memory (OOM) issue on the FE caused by an incorrect execution plan for a Union operator.

    • Fixes a query failure issue that occurs after a column is renamed.

    • Fixes an issue where the primary key index is lost due to a snapshot loading failure.

    • Fixes a refresh issue caused by an abnormal change in MV properties.

    • Uses the FE configuration item task_runs_max_history_number to control the number of historical TaskRuns saved in the information_schema.task_runs view. This reduces memory overhead.

    • Fixes an issue where versions are lost during data import.

    • Fixes an issue of abnormal Task Run information on follower nodes.

    • Fixes an issue where Broker Load cannot be resumed after an FE primary/secondary switchover.

    • Fixes an issue where a query returns null after a column is modified.

    • Fixes a partition creation failure that occurs when writing to multiple tables in a single transaction.

    • Fixes a query error caused by a CASE WHEN statement that returns an incorrect type.

    • Fixes a cache failure caused by duplicate cache keys in concurrent scenarios.

    • Fixes a partial update failure on primary key tables.

    • Fixes an issue with the lock mechanism for submitting tasks.

    • Fixes a global dictionary issue for the Union operator in low-cardinality scenarios.

    • Fixes an issue of incorrect pruning of empty partitions.

    • Fixes an error caused by the loss of null partitions during MV rewriting.

    • Fixes an issue where the splitor cannot prune scan columns.

    • Fixes issues related to Left Anti Join.

    Separation of storage and compute

    [New features]

    • Shared-data instances support full-text search.

    • Adds the distributed compaction feature to mitigate the performance bottleneck of single-node compaction.

    • Supports the adaptive I/O mode for cache to prevent performance degradation caused by I/O bottlenecks on cache disks.

    • Provides warehouse-level query metrics to help you better monitor query performance.

    [Optimized features]

    • Enables Index Cache and Metadata Cache by default to further improve query performance.

    • Fixes an issue where the minimum version information is missing for shared-data tables.

    [Fixed issues]

    • Fixes an issue where data is written abnormally to the destination warehouse during an INSERT INTO EXTERNAL TABLE operation in shared-data mode.

    • Fixes a concurrent transaction failure that occurs after Batch Publish is enabled.

    • Fixes an issue where the maximum number of compaction tasks cannot be limited after a warehouse is specified for compaction.

    • Fixes an error that occurs when prefetching data for a table using Cache Select after a schema change.

    • Fixes a failure to modify the data distribution method using an ALTER TABLE statement.

    Lakehouse

    [New feature] Paimon tables now support column-level kinship analysis and feature optimized statistics information.

    Optimizes the efficiency of retrieving DLF 2.5 metadata for Paimon queries to improve query performance.

    [Fixed issues]

    • Fixes a memory leak issue with Iceberg equality delete files.

    • Fixes a DLF 2.5 token expiration issue.

    • Fixes an issue where MVs cannot be refreshed normally after the underlying table of a foreign table is rebuilt.

    • Fixed an issue where the refresh of materialized views on external tables failed.

    • Fixes a failure to write data to Hive tables using INSERT OVERWRITE.

    • Fixes an MV refresh failure caused using the str2date function on foreign tables.

    Full-text search

    [New features]

    • Shared-data instances support full-text search.

    • Primary key tables support full-text search.

    • Added multiple Match functions to enhance full-text index capabilities.

    Optimizes the number of index files to improve access efficiency in shared-data mode.

    Version 3.3.13-1.0.0

    Release date

    June 19, 2025

    Updates

    This is the official release of the enterprise-grade kernel Stella 1.0.0. This version is based on a cloud-native, shared-data architecture. It focuses on improving system stability and performance and extends support for diverse data analytics scenarios, such as full-text search. This release aims to provide enterprises with unified and efficient cloud-native data analytics capabilities.

    Feature category

    Description

    Base engine

    [New features]

    • Supports Secure Sockets Layer (SSL) encryption.

    • Statistics collection tasks support the synchronous mode option.

    • Spark Load supports dynamic configuration updates.

    [Optimized features]

    • Optimizes the reload performance of materialized views (MVs) to significantly improve FE startup efficiency.

    • Optimizes the collaborative processing of aggregate operation pushdown and Broadcast Join.

    • Fixes an issue where the row limit does not take effect in Broadcast operations.

    • Improves the accuracy of execution plans generated by GROUP_CONCAT.

    • Optimizes the lock granularity for MV refreshes.

    • Enhances the Metrics API to provide more monitoring metrics.

    • Optimizes the timeout issue of the Get Shard Info remote procedure call (RPC) in high-concurrency scenarios.

    [Fixed issues]

    • Fixes a crash issue on CNs or BEs caused by a schema change when upgrading from version 3.2 to 3.3.

    • Fixes a data loss issue during insertion caused by Report Exec State.

    • Fixes a CN crash issue that occurs after Fast Schema Evolution is enabled.

    • Fixes an issue where an insert task does not end normally after a table is deleted.

    • Fixes an issue where a version cannot be found during a query.

    • Fixes an issue where an MV refresh task gets stuck after the task is killed.

    • Fixes an issue of incorrect execution plans generated during multi-stage MV rewriting.

    • Fixes an issue where tablets cannot be fully loaded.

    • Fixes an issue where MV relationships are not correctly built after an FE image is loaded.

    • Fixes a query failure (Query Terminated exception) caused by the DESC Table Cache.

    • Fixes an issue of abnormal transaction results.

    • Fixes an issue of incorrect results for Join and Union queries.

    • Fixes an issue of incorrect query results caused by the n_gram function.

    • Fixes an issue where kinship analysis in DataWorks does not take effect.

    • Fixes an issue of incorrect information in DESC TABLE for Paimon.

    • Fixes an issue where Paimon MVs are always fully refreshed.

    Storage and compute separation

    [New features]

    • Supports configurable Compaction Punish time.

    • Supports Tablet Meta Cache and Index Cache to significantly improve query performance.

    • The _query_plan interface supports specifying a warehouse (compute group).

    • Supports compaction scheduling across multiple warehouses.

    [Optimized features]

    • Improves the observability of compaction.

    • Uses Poco HTTP Client to optimize OSS access efficiency.

    • Supports clearing the local cache after a Drop Partition operation to avoid occupying invalid disk space.

    • Optimizes the performance of retrieving shard information.

    • Improves the direct read performance of Count operations.

    • Enhances the observability of load tasks in a multi-warehouse environment.

    • Supports adaptive file formats to reduce input/output operations per second (IOPS) and optimize query performance.

    [Fixed issues]

    • Fully enables the cloud-native primary key (PK) index and fixes several metadata-related issues.

    • Fixes an issue where compaction cannot be scheduled normally.

    Lakehouse

    [New features]

    • Supports Alibaba Cloud Data Lake Formation 2.5.

    • Paimon foreign tables support time-type fields.

    • Paimon foreign tables support querying properties using SHOW CREATE TABLE.

    [Optimized features]

    • Removes redundant I/O operations and reduces the number of I/O operations when MVs are scheduled to frequently retrieve metadata. This improves refresh efficiency.

    • Optimizes Paimon metadata cache configurations.

    • Fixes an issue where filtering on the FLOAT type does not take effect in Paimon queries.

    • Improves the performance of reading partition information from DLF.

    • Supports metadata refresh for MaxCompute foreign tables.

    • Uses UUIDs instead of creation times to optimize Paimon query performance.

    • Improves the query performance of Paimon deletion vector tables.

    • Optimizes the scanning efficiency of Parquet files.

    [Fixed issues]

    • Fixes an issue of incorrect time zones for the DATETIME type in Paimon queries.

    • Fixes a failure to read the VARCHAR type from Hudi data sources.

    • Fixes an issue where Paimon commit information is written incorrectly.

    Full-text search

    [New features]

    • The full-text search feature is released in beta. This version optimizes multiple Clucene performance issues, improving text search performance by more than 100%.

    • Supports Tokenize-related functions.

    [Optimized features]

    • Optimizes the file structure of inverted indexes to improve query efficiency.

    • Improves the Index Tool to facilitate quick problem identification and resolution.

    Version 3.3.8-1.98 and earlier

    Version number

    Release date

    Description

    3.3.8-1.98

    May 28, 2025

    [New features]

    • The multi-warehouse feature supports a compaction isolation mechanism. This allows compaction operations to be dynamically performed based on the warehouse used by the import task.

    • Paimon capabilities are enhanced. Paimon DLF 2.5 Catalog is supported. The Paimon Manifest Cache mechanism is introduced. The Paimon Schema Change feature is supported.

    [Optimized features]

    • Adds the dynamic FE configurations lake_min_compaction_interval_ms_on_success and lake_min_compaction_interval_ms_on_failure. These configurations precisely control the compaction scheduling latency in shared-data scenarios and reduce the impact of high-frequency import tasks on scheduling performance.

    • Optimization for storage-compute separation

      • Memory optimization: Introduces the Max Buffer Rows limit for partial column updates in the shared-data architecture to significantly improve instance stability.

      • gRPC protocol optimization: Supports channel reuse to significantly reduce network pressure between FEs and BEs and improve query performance.

      • RPC interface optimization: Optimizes the RPC interface from FEs to StarOS shards to further improve query performance.

      • Kernel optimization: Improves the parallel scan capability for tablets in BEs and CNs to significantly improve query efficiency.

    • Optimizes the overhead of MV reloads during FE startup and leader FE checkpoint execution. This is especially useful for clusters with many MVs and complex dependencies. It improves instance stability and reduces resource consumption.

    [Fixed issues]

    • Fixes an error caused by repeated calls to the FinishInstance RPC.

    • Fixes a Base Compaction Forbidden Range error caused by an FE checkpoint.

    • Fixes a compaction scheduling issue that occurs when the feature of binding compaction to an import warehouse is disabled.

    • Fixes a potential data loss issue in some Insert Overwrite scenarios.

    • Fixes some Paimon bugs, including time zone issues and DLF 2.5 dependency issues.

    • Fixes a CN crash issue caused by data spilling.

    • Repairing storage-compute separation:

      • Fixes a data loss issue in primary key tables with sort keys caused by partial update operations.

      • Fixes a schema change failure caused by column unique ID incompatibility after an upgrade from version 3.1 or 3.2 to 3.3.

      • Fixes a version inconsistency issue caused by Batch Publish.

    3.3.8-1.91

    March 20, 2025

    [New features]

    • Partial and conditional updates are supported for shared-data instances.

    • The dynamic configuration item enable_delete_with_condition is added for BEs. This configuration item specifies whether to allow deletion with specific conditions.

    • The configuration item enable_corrupt_check_set_bad is added for FEs. The default value is false. If you enable this item, the system checks whether replica files are damaged when you query data from tables created in shared-nothing instances. If the replica files are damaged, the system automatically marks the files as bad to improve data reliability.

    • The AccessKey ID and AccessKey secret can be explicitly configured in the statements used to create an external catalog.

    [Optimized features]

    Improves the performance of executing the SHOW MATERIALIZED VIEW statement or querying data from the information_schema.materialized_views table when many MVs exist.

    [Fixed issues]

    • Fixed an issue that prevented the use of transactional Stream Load on compute-storage integrated instances.

    • Fixes an issue where CNs crash in community version 3.3.8.

    • Fixes an issue where the physical data files of expired partitions in the recycle bin are not deleted after MVs are refreshed.

    • Fixes an issue where Kudu foreign tables cannot be accessed after you enable the multi-warehouse feature.

    • Fixes an issue where spill tasks cannot end as expected after the corresponding query context is released.

    • Fixes an issue where BEs crash when FlatJSON is used to read non-object JSON data.

    • Fixes a time zone conversion issue for timestamp fields in Paimon tables.

    • Fixes an issue where a NullPointerException occurs when the SHOW PARTITION statement is executed for non-partitioned Paimon tables.

    3.3.8-1.90

    March 04, 2025

    [New features]

    Statistics calculation and collection at the DataCache layer are supported for shared-data instances.

    [Optimized features]

    • Optimizes the logic of obtaining MVs to improve system performance.

    • Enhances compaction stability:

      • Fixes an issue where the compaction operation may be blocked after FEs are restarted.

      • Fixes an issue where the compaction operation does not automatically end.

    • Optimizes the request logic of the OSS API.

    • Provides logs and observability when you batch publish data.

    • Provides more compaction-related diagnostic analysis information, including status details and performance metrics.

    • Optimizes the access logic of OSS-HDFS to enhance the stability and efficiency of data read and write operations.

    [Fixed issues]

    • Fixes an issue where the lake compaction transaction is repeatedly triggered during FE primary/secondary switchovers.

    • Fixes an issue where the multi-warehouse feature cannot work as expected when the transactional mechanism is used with the Stream Load node.

    • Fixes an issue where the warehouse property is missing from the information_schema table.

    • Fixes an issue where the Paimon cache is not automatically refreshed.

    • Fixes an issue where Resource Access Management (RAM) user properties are lost after FEs are restarted.

    • Fixes an error that occurs during DLF 2.0 authentication.

    3.3.8-1.87

    February 06, 2025

    [New features]

    • The SHOW PARTITIONS syntax is supported for Paimon tables to view partition information.

    • Supports writing data to Paimon tables.

    • Supports DLF 2.0.

    [Optimized features]

    • Removes profile-related modifications.

    • Removes code related to Pure DLA.

    • Adds detailed observability information to compaction tasks.

    • Supports hint settings for Cache Select, which lets you control cache behavior using hints.

    • Optimizes the forced refresh policy for MVs.

    • Extends the functionality of the Update statement to support setting a column multiple times in a single statement.

    • Supports deleting the cloud-native index for shared-data tables.

    • Paimon tables support statistics collection to optimize query performance.

    • Supports Limit Push Down when a query on a Paimon table contains only partition predicates.

    • Adds a log printing feature to Metrics Action.

    • Adds primary key (PK)-related metrics to the cloud-native index.

    • Supports counting OSS API calls.

    • Optimizes the timeout issue in gRPC calls.

    • Adds scheduling logs to the compaction process.

    • Paimon tables support refreshing the manifest cache.

    • Optimizes the profile information output for Paimon table queries.

    • Shared-data tables support the index cache feature.

    • Shared-data table profiles support write-related performance metrics.

    • Optimizes MV refresh tasks to fix FE memory issues.

    • Supports the metadata cache feature for Paimon tables.

    • Supports exporting column-level kinship to DataWorks.

    • Optimizes the execution efficiency of COUNT(1) queries.

    [Fixed issues]

    • Fixes the thrift class not found exception that occurs when querying DLF Paimon tables.

    • Fixes an exception caused by type conversion during Paimon table queries.

    • Fixes an issue where the aliyun.oss.endpoint configuration cannot be correctly identified in Paimon catalog queries.

    • Fixes an issue of inaccurate results for aggregate queries.

    • Fixes an issue of incorrect statistics collection results for data lake tables.

    • Fixes an issue of abnormal segment files generated after a partial update operation is performed on internal shared-data tables.

    • Fixes an issue where the compaction profile is displayed incorrectly.

    • Fixes an issue where the LockManager of FEs fails to wake up all callers.

    • Fixes an issue where sort keys cannot be modified.

    • Fixes an issue where the warehouse property configuration does not take effect.

    • Fixes an issue where the admin user cannot manage warehouses.

    • Fixes an issue of abnormal Reduce rules in the cost-based optimizer (CBO).

    • Fixes an issue where an error message indicating that a Hudi table does not exist is incorrectly reported when querying a Hive table.

    • Fixes an issue where the publish time is not correctly updated after an update operation on a shared-data table.

    • Fixes an error that occurs when reading Parquet files written by Spark.

    3.3.2-1.77

    November 19, 2024

    • Supports cloud-native PK indexes.

    • Supports metric statistics for OSS API access for cost optimization.

    • Optimizes the fault tolerance of FE query scheduling to improve stability.

    • Fixes a stability issue in community version 3.3 caused by Prepare log recording.

    • Enables Paimon catalog-level cache by default to improve the read speed of Paimon tables.

    • Paimon tables support performance analysis using EXPLAIN ANALYZE.

    • Optimizes execution plans for Paimon tables.

    • Fixes a potential deadlock issue for Paimon tables in a Hive or DLF catalog.

    • Fixes an issue where Paimon system tables cannot be queried.

    • Fixes an issue where DLF foreign tables cannot be accessed.

    3.2

    Note

    EMR Serverless StarRocks 3.2 is based on the Linux Foundation Project StarRocks. For more information, see StarRocks version 3.2.

    Version number

    Release date

    Description

    3.2.15-1.97

    May 28, 2025

    [Fixed issues]

    • Fixes an error that might occur when parsing properties in an Elasticsearch catalog.

    • Fixes an issue where the query timeout cannot be correctly set during the execution of an Optimize Job.

    • Fixes an issue where a CreateLoadTask operation might cause an FE node to get stuck.

    • Fixes an issue where Stream Load reports a duplicate label due to an abnormal transaction status.

    3.2.15-1.92

    March 21, 2025

    [New features]

    Supports the MAX_BY() and MIN_BY() window functions.

    [Optimized features]

    • Supports setting a time range during which base compaction is prohibited for a table.

    • Supports collecting statistics on Paimon tables.

    • Adds node and histogram metrics to Metrics.

    • Adds timeout parameter configurations in StarClient.

    • Supports automatic detection of damaged replicas and marks them as bad.

    [Fixed issues]

    • Fixes an issue where 0 is returned for the loadRowsRate field after SHOW ROUTINE LOAD is executed.

    • Fixes an issue where the Files() function cannot correctly read unqueried columns.

    • Fixes an issue where the ARRAY_MAP function causes BEs to crash.

    • Fixes an issue where Metadata Cache causes BEs to crash.

    • Fixes an issue where Routine Load is canceled due to a transaction timeout.

    • Fixes an issue where the execution plan for the nested aggregate function MAX(COUNT(DISTINCT)) is incorrect when low-cardinality optimization is enabled.

    • Fixes an issue where Stream Load is redirected to non-alive nodes.

    • Fixes an issue where bRPC continuously reports errors after a BE is restarted.

    • Fixes an issue where Stream Load fails to be submitted using HTTP 1.0.

    • Fixes an issue where the task status is inconsistent between leader and follower FEs.

    • Fixes an issue of an incorrect row count in MVs.

    3.2.11-1.79

    November 20, 2024

    [Optimized features]

    • Enables Paimon partition-level cache by default to improve query speed.

    • Optimizes the statistics feature of Paimon.

    • Adds error messages for failed queries to the audit log.

    [Fixed issues]

    • Fixes an issue of frequent migration of colocate tablets.

    • Fixes an issue where the OSS scheme cannot be found when accessing Paimon tables on OSS or OSS-HDFS.

    • Fixes an issue where DLF 2.0 Paimon tables cannot be created.

    • Fixes a partition error that occurs when writing data to Hive partitioned tables.

    • Fixes an error when the array_to_bitmap function processes constant results.

    3.2.11-1.76

    October 30, 2024

    [Optimized features]

    • Optimizes profile collection policies and fixes potential crash issues.

    • Enhances the exception handling mechanism of the shared-data architecture to improve system fault tolerance.

    [Fixed issues]

    • Fixes an issue of incorrect results when inserting data into external partitioned tables.

    • Fixes an issue of abnormal FEs caused by thread leaks.

    • Fixes a crash issue that occurs when PK tables use the Bitmap type in a shared-data environment.

    • Fixes a ToDataCacheInfo leak issue to prevent OOM errors on FEs in the shared-data architecture.

    • Fixes an issue of incorrect query results caused by Query Cache.

    • Fixes an issue where some Bucket Shuffle operations return incorrect results.

    • Fixes an issue where a NullPointerException (NPE) is generated when accessing table functions, which improves system stability.

    3.2.9-1.71

    September 14, 2024

    [New features]

    • Supports writing data to Paimon tables.

    • Supports reading from and writing to DLF 2.0 Paimon catalogs.

    [Optimized features]

    • Supports the Explain Analyze feature for Paimon tables.

    • Removes pruning and statistics for Paimon tables.

    • Adds an SQL field to Current Query.

    • Optimizes the performance of SELECT COUNT.

    • Adds an interface to obtain migration progress.

    • Iceberg DLF supports configuring dlf.catalog.id.

    [Fixed issues]

    • Fixes a LIKE query rewrite exception.

    • Fixes an issue where `show data` displays inaccurate data.

    3.2.9-1.67

    August 16, 2024

    [Optimized features]

    Optimizes the performance of Hive Sink.

    [Fixed issues]

    • Fixes a permission adaptation issue between Hive Catalog and Ranger.

    • Fixes a crash issue caused by Count optimization.

    • Fixes an issue of low MV refresh efficiency caused by obtaining snapinfo.

    • Fixes an issue of metadata expansion in the Compaction Manager for shared-data tables.

    • Fixes an issue that causes CNs to crash.

    • Fixes an issue where Prepare Statement cannot be used after an upgrade to 3.2.9.

    • Fixes a data inconsistency issue that occurs after data spilling is enabled.

    3.2.9-1.66

    August 09, 2024

    [New features]

    • Supports writing to OSS-HDFS files when the parent directory does not exist.

    • Broker Load supports Jindo.

    • Implements automatic identification of different engines for OSS-HDFS.

    [Optimized features]

    • Optimizes compaction logs for more efficient diagnosis and analysis.

    • Creates the star mgr directory when CNs start.

    • Optimizes the default redirection configurations of FEs.

    • Optimizes unnecessary log records and configuration items to improve performance.

    • Adds more I/O monitoring metrics for the shared-data architecture to improve overall O&M capabilities.

    • Optimizes the adaptive I/O policy for the shared-data architecture.

    [Fixed issues]

    • Supports accessing GCS and Azure data using foreign tables.

    • Fixes an error that occurs when the spill result exceeds 4 GB.

    • Fixes an issue of incorrect results caused by spill pre-aggregation.

    • Fixes an issue of unexpected MV refreshes.

    • Fixes a PK index memory statistics issue.

    • Fixes an issue of inaccurate statistics information.

    • Fixes an issue where insert tasks cannot be canceled.

    • Fixes a plan error caused by a schema change.

    • Fixed an issue where external tables failed to retrieve CN information.

    3.2.9-1.65

    July 19, 2024

    [New features]

    • Upgrades the Jindo software development kit (SDK) to version 6.5.0.

    • Iceberg supports DLF Catalog.

    • Shared-data clusters support shard rebalancing.

    • Supports the regexp_split method.

    [Optimized features]

    • The default wait time of get_txn_status_internal_sec is shortened from 30 seconds to 10 seconds to prevent Stream Load from being delayed in the publish phase.

    • Supports dynamic modification of the pindex_shared_data_gc_evict_interval_seconds parameter to adjust the garbage collection (GC) interval of the local persistent index for primary key tables in shared-data clusters.

    [Fixed issues]

    • Fixes an occasional Jindo CRC mismatch issue when exporting data to OSS.

    • Fixes an occasional issue where a file does not exist when accessing OSS-HDFS.

    • Fixes an issue where an SQL syntax error is misreported as "Lost Connection".

    3.2.8-1.62

    June 27, 2024

    [New features]

    • Unified Catalog supports Kudu and Paimon, and supports creating DLF-type Unified Catalogs.

    • Supports generating kinship logs.

    • Supports viewing comment information of foreign tables when using the describe and show create commands.

    [Fixed issues]

    • Fixes an issue where data cannot be written to foreign tables in shared-data clusters.

    • Fixes a potential refresh failure of Paimon MVs if a partition key column contains NULL values.

    • Fixes an issue of inaccurate memory statistics.

    3.2.6-1.60

    June 06, 2024

    [New features]

    • Supports categorizing SQL statements that fail to be parsed as "bad sql".

    • Kudu connectors are supported.

    Adds the enable_pipeline_engine configuration.

    [Fixed issues]

    • Fixes an issue where a cluster migration task fails to correctly obtain the source cluster schema, causing BEs to crash.

    • Fixes an issue of inaccurate memory statistics.

    • Fixes a performance issue when Paimon reads from read-only tables.

    3.2.6-1.59

    May 31, 2024

    [New Feature] Displays "bad sql" information in Query Details for syntax parsing errors.

    Supports delete vector queries for Paimon.

    [Fixed issues]

    • Fixes an issue where Paimon Catalog cannot use custom DLF directories.

    • Fixes an issue where a warehouse must be specified when creating a Paimon Catalog.

    3.2.6-1.57

    May 23, 2024

    Note

    If you use a version earlier than 3.2.6-1.57, we recommend that you upgrade to 3.2.6-1.57 or a later version.

    Supports optimizing the bucket storage layout of internal shared-data tables using the OPTIMIZE command.

    [Optimized features]

    • Adds the FE configuration fragment_profile_drop_threshold_ms, which supports default configurations in the console. The default threshold for drop fragment profile is set to 0, which means this feature is disabled by default.

    • Provides an independent thread pool configuration lake_flush_thread_num_per_store for refreshing internal shared-data tables. The default value is 2 * Number of CPUs. This is intended to optimize write I/O throughput performance.

    • Optimizes the I/O merge policy to allow direct reading of entire small files.

    • Adjusts the default values of some cross-cluster migration configurations.

    [Fixed issues]

    • Fixes an issue where a connection to the report_exec_stat RPC fails and cannot be re-established, which causes INSERT INTO operations and profile collection to fail.

    • Fixes an issue introduced by MV Index that causes FEs to crash frequently during metadata replay.

    • Fixes an AST Cache issue introduced after version 3.2.6 that prevents some MVs from being created.

    3.2.6-1.52

    May 08, 2024

    Supports creating DLF database and table metadata.

    [Fixed issues]

    • Fixes an error in obtaining tablet metadata after the lake_tablet_internal_parallel configuration is enabled.

    • Fixes a timeout issue for INSERT INTO import operations. The thrift_rpc_timeout_ms configuration is added to BEs to adjust the RPC timeout.

    • Fixes an issue where a schema change does not respond for a long time.

    3.2.4-1.37

    March 08, 2024

    • Supports cross-cluster data migration.

    • Primary key (PK) tables support size-tiered compaction.

    References

    For information about the latest feature updates of the EMR Serverless StarRocks console, see Console release notes.