V4.3.5 documentation updates|V4.3.5| docs|Distributed Database

V4.3.5 documentation updates

Last Updated：2025-11-18 14:33:32 Updated

This topic mainly introduces the update records for parameters, system variables, views, functions, syntax, and data types in OceanBase Database V4.3.5 BP versions. For more information about feature changes, see OceanBase Database V4.3.5.

V4.3.5 BP4 (2025-09-10)

Feature	Description
New parameter `enable_mlog_auto_maintenance`	Controls whether to enable the automatic creation of MLOGs. For more information, see enable_mlog_auto_maintenance.
New parameter `mlog_trim_interval`	Controls the scheduling cycle for the background automatic trimming task of MLOGs. For more information, see mlog_trim_interval.
New parameter `ob_deterministic_udf_cache_max_size`	Specifies the maximum total size of the cache pool for deterministic UDFs during statement execution, in bytes. If the value is 0, the execution-time cache feature is disabled.
New parameter `approx_count_distinct_precision`	Specifies the precision of the result of the `approx_count_distinct` function. The default value is 10.
New parameter `enable_ps_parameterize`	Controls whether to enable parameterization of PS statements during parsing.
Changed the behavior of the `ob_route_policy` parameter	`FORCE_READONLY_ZONE`: After the parameter value is set, non-weak read requests (strong read/DML) are executed normally without errors. `COLUMN_STORE_ONLY`: After the parameter value is set, non-weak read requests (strong read/DML) are executed normally without errors. For more information, see ob_route_policy.
Added columns to the view `[G]V$OB_SQL_AUDIT`	Added the `stmt_type` column to identify the SQL type for statistical analysis. Added the `trans_status` column to describe whether the transaction is explicitly or implicitly started. Added the `user_client_port` column to display the user client port number. In a regular tenant, the `query_sql` column now displays NULL values. For more information, see GV$OB_SQL_AUDIT.
Changed the meaning of the `row_movement` field in the views `CDB/DBA/USER/ALL_TABLES`	The `row_movement` field now records the actual enable/disable status of row movement instead of being NULL. For more information, see ALL_TABLES.
Changed the meaning of the `row_movement` field in the view `DBA/USER/ALL_ALL_TABLES`	The `row_movement` field now records the actual enable/disable status of row movement instead of being NULL. For more information, see ALL_ALL_TABLES.
`ALTER TABLE`	In MySQL-compatible mode, the `ALTER TABLE ADD INDEX` syntax mocks the `ALGORITHM` and `LOCK` attributes. For more information, see ALTER TABLE.
`UTL_RAW.TRANSLATE`	In Oracle-compatible mode, the `UTL_RAW.TRANSLATE` function is added. This function translates data of the `RAW` type byte by byte.
`CREATE MATERIALIZED VIEW`	When creating a materialized view, adding `AS OF PROCTIME()` to the base table allows you to specify that the table is skipped during incremental refreshes. Additionally, tables with the `AS OF PROCTIME()` clause can have MLOGs created. For more information, see CREATE MATERIALIZED VIEW (MySQL-compatible mode) and CREATE MATERIALIZED VIEW (Oracle-compatible mode).
`ALTER SYSTEM LOAD MODULE DATA MODULE`	Added the syntax for importing specified tenant data, including Redis, timezone metadata, and spatial data srsid. This command can only be executed by the sys tenant.

V4.3.5 BP3 (2025-07-21)

Feature	Description
Changed the default value of the `ob_vector_memory_limit_percentage` parameter	The original default value was 0, indicating that no memory was allocated for vector indexes, and creating an index would result in an error. The default value is now 0, indicating adaptive memory allocation for vector retrieval. For more information, see ob_vector_memory_limit_percentage.
Changed the default value of the `auto_split_tablet_size` parameter	The default value has been changed from 128 MB to 2 GB. For more information, see auto_split_tablet_size.
New parameter `default_table_merge_engine`	Controls the default `merge_engine` for table creation. For more information, see default_table_merge_engine.
New view `[G]V$OB_VECTOR_MEMORY`	Displays memory usage statistics for the tenant's vector index module.
Added the `TOP_TIME` column to the `[G]V$OB_PROCESSLIST` view	The `TOP_TIME` column indicates the total execution time of the PL. For more information, see GV$OB_PROCESSLIST.
Added the `EXECUTE_TIME` column to the `GV$OB_TABLET_COMPACTION_HISTORY` view	The `EXECUTE_TIME` column indicates the CPU consumption time for a major compaction. For more information, see GV$OB_TABLET_COMPACTION_HISTORY.
Added parameters to the `DBMS_MVIEW.REFRESH` system package	Set the refresh strategy for nested materialized views. When `nested` is set to `true`, nested materialized views are cascaded refreshed. When `nested` is set to `false`, nested materialized views are independently refreshed. Set the consistency behavior for cascading refreshes. When `nested` is set to `true`, specify `consistent` and `inconsistent` to describe the two cascading refresh behaviors. If the `nested_refresh_mode` keyword is not specified, the default is cascading non-consistent refresh. For more information, see REFRESH (MySQL-compatible mode) or REFRESH (Oracle-compatible mode).
`CREATE MATERIALIZED VIEW`	When creating a nested materialized view, you can specify the refresh strategy: `individual` indicates independent refresh. `inconsistent` indicates cascading non-consistent refresh. `consistent` indicates cascading consistent refresh. For more information, see CREATE MATERIALIZED VIEW.
`ALTER MATERIALIZED VIEW`	You can modify the refresh strategy of a nested materialized view. The semantics of the three keywords are consistent with those of the `CREATE MATERIALIZED VIEW` statement. For more information, see ALTER MATERIALIZED VIEW.
`CREATE MATERIALIZED VIEW LOG`	Add the `OR REPLACE` clause to the `CREATE MATERIALIZED VIEW LOG` statement. For more information, see CREATE MATERIALIZED VIEW LOG.
`CREATE VECTOR INDEX`	Add the `nbits` parameter to the IVFPQ vector construction parameters. `nbits` indicates the number of pq cluster centers, which is 2 raised to the power of `nbits`. The compressed size of the pq code is `nbits`. Change the maximum value of the `nlist` parameter in the IVF vector construction parameters from 65535 to 65536. Change the range of the `m` parameter in the IVFPQ vector construction parameters from m > 0 to [1, 65536]. The default value is dim/8. For more information, see Create a vector index.

V4.3.5 BP2 (2025-05-15)

Feature	Description
New parameter `default_storage_cache_policy`	This parameter is used to configure the default cache policy. For more information, see default_storage_cache_policy.
New parameter `enable_manual_storage_cache_policy`	This parameter is used to enable the hot cache strategy specified by users. For more information, see enable_manual_storage_cache_policy.
New parameter `suspend_storage_cache_task`	This parameter specifies whether to suspend the background cache task. For more information, see suspend_storage_cache_task.
New parameter `enable_lock_priority`	This parameter is used to control the table lock priority. For more information, see enable_lock_priority.
New parameter `enable_adaptive_plan_cache`	This parameter is used to enable or disable the adaptive plan cache feature for a tenant.
New parameter `clog_io_isolation_mode`	Specifies whether to use resource isolation to control I/O requests on the CLOG disk, depending on whether the CLOG disk and the data disk are isolated. For more information, see clog_io_isolation_mode.
New parameter `global_index_auto_split_policy`	Specifies whether to enable automatic partitioning for newly created global indexes. For more information, see global_index_auto_split_policy.
New system variable `event_scheduler`	This variable determines whether the event scheduler is enabled and running on the server. For more information, see event_scheduler.
New system variable `ob_enable_ps_parameter_anonymous_block`	This variable specifies whether to parameterize anonymous blocks and `call` statements in parsing in PS mode.
New system variable `ob_sparse_drop_ratio_search`	Specifies the percentage of values that can be ignored in all dimensions of a vector during queries.
New view `information_schema.events`	This view records event information. For more information, see information_schema.events.
New view `[G]V$OB_STORAGE_CACHE_TASKS`	Displays information about storage cache tasks. For more information, see GV$OB_STORAGE_CACHE_TASKS.
New view `[G]V$OB_TABLET_LOCAL_CACHE`	Display information about tablet local cache. For more information, see GV$OB_TABLET_LOCAL_CACHE.
New view `V$OB_DYNAMIC_PARTITION_TABLES`	Displays information about dynamic partition tables. For more information, see V$OB_DYNAMIC_PARTITION_TABLES.
New view `CDB/DBA_OB_DYNAMIC_PARTITION_TABLES`	This view is used to query information about dynamic partition tables. For more information, see oceanbase.CDB_OB_DYNAMIC_PARTITION_TABLES.
New view `CDB/DBA_MVIEW_RUNNING_JOBS`	This view displays the current running jobs related to materialized views, including refresh jobs, MLOG cleanup jobs, and so on.
New view `DBA_MVIEW_DEPS`	This view displays the dependencies of a materialized view.
Added columns `ss_data_disk_operation_suggested` and `ss_data_disk_size_suggested` to the `[G]V$OB_SERVERS` view.	The new columns are used to display information about suggesting operations for the local SSD disks of shared storage servers, such as the size of the local SSD disks. For more information, see GV$OB_SERVERS.
Added columns to view `[G]V$OB_TABLET_COMPACTION_HISTORY`.	A new column `mds_filter_info` is added, which specifies the MDS filtering information used for minor compaction. For more information, see GV$OB_TABLET_COMPACTION_HISTORY.
Added columns to view `[G]V$OB_SQL_AUDIT`.	The `INSERT_DUPLICATE_ROW_COUNT` column is added to indicate the number of conflicting rows in the `INSERTUP` or `REPLACE INTO` statements. For more information, see GV$OB_SQL_AUDIT.
Added columns to views `CDB/DBA/ALL/USER_MVIEWS`.	The `DATA_SYNC_SCN` and `DATA_SYNC_DELAY` columns are added. For nonnested materialized views (views that depend on only base tables), `data_sync_scn` indicates the refresh point. For nested materialized views, `data_sync_scn` is the smallest `data_sync_scn` value of the materialized views that the materialized view depends on. `data_sync_delay` is a calculated indicator based on the `data_sync_scn` value for user reference. `data_sync_delay = current_time - data_sync_scn`. `DATA_SYNC_DELAY` is merely a reference value to the amount of data delay in the materialized view. For example, `data_sync_delay` is not zero if the base table is not refreshed from the materialized view after the materialized view was last refreshed, which indicates that the materialized view data has not been delayed. For more information, see CDB_MVIEWS.
Added columns to views `CDB/DBA/ALL/USER_MVIEW_LOGS`	The `PURGE_DOP` and `LAST_PURGE_TIME` columns are added to record the parallelism and execution time of the previous purge task on a materialized view log. For more information, see CDB_MVIEW_LOGS.
Added columns to views `DBA/USER_AUDIT_SESSION`	The `svr_ip` and `svr_port` columns are added to identify Server Sessions based on the same Client Session ID on different OBServer nodes. For more information, see DBA_AUDIT_SESSION.
Added columns to views `CDB/DBA_WR_ACTIVE_SESSION_HISTORY`.	The `DELTA_READ_IO_REQUESTS`, `DELTA_READ_IO_BYTES`, `DELTA_WRITE_IO_REQUESTS`, and `DELTA_WRITE_IO_BYTES` columns were added. For more information, see oceanbase.CDB_WR_ACTIVE_SESSION_HISTORY.
Added columns to view `[G]V$ACTIVE_SESSION_HISTORY`.	The `DELTA_READ_IO_REQUESTS`, `DELTA_READ_IO_BYTES`, `DELTA_WRITE_IO_REQUESTS`, and `DELTA_WRITE_IO_BYTES` columns were added. For more information, see GV$ACTIVE_SESSION_HISTORY.
Added columns to views `[G]V$OB_ACTIVE_SESSION_HISTORY`.	The `DELTA_READ_IO_REQUESTS`, `DELTA_READ_IO_BYTES`, `DELTA_WRITE_IO_REQUESTS`, and `DELTA_WRITE_IO_BYTES` columns are added. For more information, see GV$OB_ACTIVE_SESSION_HISTORY.
Changed output of the `information_schema.referential_constraints`.	The values of the `update_rule` and `delete_rule` columns are updated from the "SET_NULL" and "NO_ACTION" values in the MySQL enum format to the "SET NULL" and "NO ACTION" values in the MySQL compatible format. For more information, see information_schema.referential_constraints.
Changed semantics of the `SESSION_ID` field.	The semantic definition of the `SESSION_ID` field has changed in V4.3.5 BP2 as follows: In ODP mode: If the configuration parameter `client_session_id_version = 2` is set in ODP, this field indicates the Client Session ID. If `client_session_id_version = 1` is set, this field indicates the Server Session ID.
Refactored the `pivot`/`unpivot` function	Refactored the complete pipeline from parsing to execution of the `pivot`/`unpivot` function to support use in combination with complex joins, rework the execution logic of the unpivot operator, and support Vectorization 2.0.
New function `rb_or_cardinality_agg(roaringbitmap)`	The function is used to calculate the cardinality of the union of the roaringbitmap column. For more information, see rb_or_cardinality_agg.
New function `rb_and_cardinality_agg(roaringbitmap)`	This function calculates the cardinality of the intersection of the `roaringbitmap` column. For more information, see rb_and_cardinality_agg.
New function `rb_to_array(roaringbitmap)`	Outputs an ordered array of integers corresponding to the roaringbitmap. For more information, see rb_to_array.
New function `rb_contains(roaringbitmap1, roaringbitmap2)`	Determines whether roaringbitmap1 completely contains roaringbitmap2. For more information, see rb_contains.
New function `rb_contains(roaringbitmap, offset)`	This function determines whether a roaring bitmap contains a specific offset. For more information, see rb_contains.
New function `map_keys`	Batch extract all keys from a map and return the result as an array. For more information, see map_keys.
New function `map_values`	It extracts the values of all keys in a map and returns the results as an array. For more information, see map_values.
New system package `DBMS_PARTITION`	Provides a dynamic partition management interface. For more information, see Overview of DBMS_PARTITION (MySQL-compatible mode) and Overview of DBMS_PARTITION (Oracle-compatible mode).
New backup and archive configuration commands for OSS WORM	The `enable_worm` parameter is supported. For more information, see SET DATA_BACKUP_DEST and SET LOG_ARCHIVE_DEST.
New command for dynamically updating ak/sk of backups	Changes the `access_id` and `access_key` of the external storage destination. For more information, see ALTER/CHANGE/MODIFY EXTERNAL_STORAGE_DEST.
New event management commands in MySQL-compatible mode	The `CREATE EVENT` statement is used to create and schedule a new event. The `ALTER EVENT` statement is used to modify an existing event. The `DROP EVENT` statement is used to delete an event.
New pseudo column names for partitioning	Allows you to obtain the ID of the partition in which each row of data is stored.
New table property `enable_macro_block_bloom_filter`	The value indicates whether to persist bloom filters at the macroblock level. For more information, see CREATE TABLE and ALTER TABLE.
New table property `dynamic_partition_policy`	You can specify the dynamic partition management attribute of the table for automatic partition creation and drop. For more information, see CREATE TABLE and ALTER TABLE.
New table property `semistruct_encoding_type`	Specifies the encoding type of the semistructured data. For more information, see CREATE TABLE and ALTER TABLE.
New catalog syntax in MySQL-compatible mode	`CREATE EXTERNAL CATALOG` creates an external data catalog. For more information, see CREATE EXTERNAL CATALOG. `SET CATALOG` changes the current catalog. For more information, see SET CATALOG. `DROP CATALOG` drops an external data catalog. For more information, see DROP CATALOG.
New `LOG ERRORS` syntax for loading data into external tables via URL	Enables error diagnostics during the external table import via URL. For more information, see LOAD DATA.
Added the `NGRAM2` tokenizer and the related `PARSER_PROPERTIES` configuration command	Specifies the number of contiguous characters to split into. For more information, see CREATE INDEX.
New `extra_info_max_size` property for creating HNSW vector indexes	A build parameter for HNSW indexes used to set the maximum number of leaf nodes per node. For more information, see Create a vector index.
New HNSW_BQ vector index creation syntax	The recall rate of HNSW_BQ indexes is slightly lower than that of HNSW indexes, but memory usage is significantly reduced. For more information, see Create vector indexes.

V4.3.5 BP1 (2025-03-18)

Feature	Description
New parameter `default_table_organization`	Allows users to set the default table organization mode when creating new tables. For more information, see default_table_organization.
New parameter `px_node_policy`	Determines the candidate pool for PX compute nodes.
New parameter `ob_enable_java_env`	Controls whether to enable the Java environment required by external table capabilities such as HDFS/MaxCompute (ODPS).
New parameter `ob_java_home`	Sets the Java home path required by JNI runtime dependencies. Must be configured when `ob_enable_java_env` is enabled.
New parameter `ob_java_opts`	Sets JVM startup parameters such as memory size and garbage collection algorithms. Must be configured when `ob_enable_java_env` is enabled.
New parameter `ob_java_connector_path`	Sets the path for the jar files required by external table connectors. Must be configured when `ob_enable_java_env` is enabled.
New parameter `temporary_file_max_disk_size`	Limits the maximum disk space that temporary files can occupy on a single node.
New parameter `package_state_sync_max_size`	Controls the maximum length of package variable values synchronized in package state. Variables exceeding this parameter will not be synchronized.
New parameter `unit_gc_wait_time`	Specifies the maximum wait time for Unit GC; if exceeded, GC is forced. For more information, see unit_gc_wait_time.
Default value change for parameter `kv_group_commit_batch_size`	The default value has been changed from 1 to 10. Group commit is now enabled by default, and the `batch_size` for group commit is set to 10. For more information, see kv_group_commit_batch_size.
New system variable `ob_ivf_nprobes`	Sets the number of nearest cluster centers to search for any IVF vector index queries in the session. A larger value explores more cluster centers, increasing search accuracy but also search time. For more information, see ob_ivf_nprobes.
New system variable `mview_refresh_dop`	Controls the default parallelism for materialized view refreshes in the session. For more information, see mview_refresh_dop.
New system variable `ENABLE_OPTIMIZER_ROWGOAL`	Controls whether the optimizer considers row goals (such as those from `LIMIT` operators) during cardinality estimation.
New system variable `pid_file`	Queries the pathname of the file where the server writes its process ID. For more information, see pid_file.
New system variable `port`	Sets the TCP/IP port number that the server listens to when deployed on an OBServer node. For more information, see port.
New system variable `socket`	Queries the name of the local client connection socket file. For more information, see socket.
New value for system variable `ob_route_policy`	A new value `FORCE_READONLY_ZONE` has been added. For more information, see ob_route_policy.
New view `CDB/DBA_OB_CS_REPLICA_STATS`	Displays the progress and status of log streams being converted to column store on each column store replica. For more information, see oceanbase.CDB_OB_CS_REPLICA_STATS.
New column added to view `information_schema.TABLES`	Adds column `ORGANIZATION`, which indicates the table organization mode. For more information, see information_schema.TABLES.
New column added to view `[G]V$OB_GROUP_IO_STAT`	Adds column `NORM_IOPS`, which displays the real-time IOPS status for each `group` under a tenant. For more information, see GV$OB_GROUP_IO_STAT.
Column value change in view `CDB_TABLES`	Changes the value of column `IOT_TYPE` from ineffective to displaying the table organization mode. For more information, see oceanbase.CDB_TABLES.
Column value change in view `[G]V$SQL_PLAN_MONITOR`	Changes the values of columns `SQL_ID` and `SQL_PLAN_HASH_VALUE` from ineffective to recording the actual values. For more information, see GV$SQL_PLAN_MONITOR.
New column and semantic change in view `[G]V$OB_SQL_PLAN`	Adds column `tenant_id` to indicate the tenant to which the plan information belongs. For the system tenant, changes query behavior from querying only its own plan information to querying plan information for all tenants. For more information, see GV$OB_SQL_PLAN.
New function `TO_CHAR` in MySQL-compatible mode	Converts other data types to string type.
New function `KEYVALUE` in MySQL-compatible mode	Parses key-value pair strings stored in data tables and extracts the value for a given key.
New function `URL_DECODE` in MySQL-compatible mode	Converts strings from `application/x-www-form-urlencoded MIME` format to regular characters.
New function `SPLIT_PART` in MySQL-compatible mode	Splits string `str` according to the separator `separator`, and returns the substring from the `start` part to the `end` part.
New function `PERCENTILE_CONT` in MySQL-compatible mode	Returns an interpolated value at the given percentile for a specified sort specification.
New system package `UTL_TCP` in Oracle mode	Supports communication with external servers via the TCP/IP protocol. For more information, see UTL_TCP Overview.
New system package `UTL_SMTP` in Oracle mode	Supports sending emails via SMTP (Simple Mail Transfer Protocol). For more information, see UTL_SMTP Overview.
New system package `DBMS_DDL` in Oracle mode	Supports encryption of PL/SQL source code.
New subprogram `DBMS_UTILITY.COMMA_TO_TABLE` in Oracle mode	Adds a subprogram to `DBMS_UTILITY` that converts a comma-separated string into a PL/SQL table. For more information, see COMMA_TO_TABLE.
New `ORGANIZATION` table option	Specifies whether a newly created table is a heap-organized table or an index-organized table. For more information, see CREATE TABLE.
New command for recompiling procedures/functions	Adds `ALTER PROCEDURE/FUNCTION COMPILE` syntax for manual compilation in MySQL tenants. The `ALTER COMPILE` statement syntax is: `alter {procedure\|function} name compile [reuse settings]`. With the `reuse settings` clause, compiles using the original PL object settings. Without the clause, compiles using the current session settings. For more information, see ALTER COMPILE.
Supports commands for changing replicated table attributes	Converts a regular table to a replicated table: `ALTER TABLE t1 DUPLICATE_SCOPE = 'CLUSTER';`. Converts a replicated table to a regular table: `ALTER TABLE t1 DUPLICATE_SCOPE = 'NONE';`. For more information, see ALTER TABLE (MySQL mode) or ALTER TABLE (Oracle mode).
New syntax for URL external tables	Adds `SELECT FROM` URL syntax. Adds `FORMAT`/`PROPERTIES`/`PATTERN` syntax to `LOAD DATA`. For more information, see SELECT (MySQL mode) or SIMPLE SELECT (Oracle mode).
New syntax for adding/removing secondary partitions in MySQL mode	Extends `RANGE [COLUMNS]`/`LIST [COLUMNS]` secondary partition add/remove syntax in MySQL mode. For more information, see Add Partition.
New command for modifying materialized views	Adds `ALTER MATERIALIZED VIEW`/`ALTER MATERIALIZED VIEW LOG` syntax to support modifying materialized view and log attributes. For more information, see ALTER MATERIALIZED VIEW and ALTER MATERIALIZED VIEW LOG.
New full-text index parser property	Adds a parser property for full-text indexes, supporting table-level `PARSER_PROPERTIES` configuration. For more information, see CREATE INDEX.
New syntax for vector indexes	Supports creating `IVF` and `HNSW_SQ` indexes with tables. The `type` parameter now supports `hnsw_sq` and `ivf_flat`, and new parameters `nlist` and `sample_per_nlist` specify the number of cluster centers and the sample size per center. For more information, see Vector Index.
New syntax for GROUPING SETS, CUBE, QUALIFY in MySQL mode	Adds `GROUPING SETS`, `CUBE`, and `QUALIFY` syntax in MySQL mode. For more information, see SELECT.
New `STRING` data type in MySQL mode	Adds a new `STRING` data type to better support AP workloads. No need to specify length, with a default maximum of 16MB. When column length is less than 16KB and not greater than `lob_inrow_threshold`, supports use as a primary key or index key. For more information, see STRING Type.
New character sets	Adds several Asian character sets and collations, such as `ujis`, `euckr`, `gb2312`, `cp932`, `eucjpms`, `cp850`, `hp8`, `macroman`, `swe7` and related collations. For more information, see Character Sets (MySQL mode) or Character Sets and Collations (Oracle mode).