New Features in Version 8.1.3

Partition addition and partition table management are supported.

PERIOD (interval type) and TTL (interval type) are now available in partition management.

CREATE TABLE

ALTER TABLE

CREATE TABLE PARTITION

JSON foreign tables are supported.

1. READ ONLY HDFS foreign tables in JSON format are supported.
2. READ ONLY OBS foreign tables in JSON file format are supported.

CREATE FOREIGN TABLE (SQL on Hadoop or OBS)

Cross-Cluster interconnection

1. Constraints are removed:
   1. The hang constraint of GDS is removed.
   2. The constraints on common aggregation functions, such as sum, count, min, max, and avg are removed.
   3. The constraints on scalar functions and analytical functions are removed.
   4. The strong verification of the column names and column types of foreign tables and source tables is added.
   5. The constraint that the column name cannot contain WHERE is removed.
   6. The limit constraint is removed.
   7. The temporary GDS foreign table policy is improved.
2. If the local cluster is the destination cluster, you can run the following statements:

GDS-based Cross-Cluster Interconnection

Enhanced cold and hot table functions

Automatic migration of cold and hot tables

CREATE TABLE

Enhanced stream data warehouse functions

The time series database stores stream data in real time.

Stream Data Warehouse GUC Parameters

CREATE INDEX

Functions and Expressions

The row-store ring buffer is controlled by a switch.

The ring buffer is controlled by a switch and its threshold is adjustable. Batch update of large tables no longer uses the ring buffer.

buffer_ring_ratio

Default storage mode of a table

It is controlled by the parameter default_orientation. If the storage mode is not specified during table creation, this parameter specifies the default storage mode.

default_orientation

GDS supports the euro sign.

The GBK character set supports the euro sign.

CREATE DATABASE

OBS foreign tables support XSKY S3.

OBS foreign tables are adapted to XSKY and the parameter eol is added.

CREATE FOREIGN TABLE (SQL on Hadoop or OBS)

SQL supports Roaring Bitmaps.

In the Internet, new retail, education, and gaming industries, target crowds need to be quickly selected based on their profiles. The SQL syntax now supports Roaring Bitmaps, which can be used for user profiling in Internet operations.

Bitmap Functions and Operators

RoaringBitmap

Case-insensitive is supported to improve Teradata compatibility.

In the data migration from Teradata to DWS, the source data contains a large amount of duplicate data with different letter cases. If DWS uses the distinct syntax to deduplicate the data, after deduplication, there is still a large amount of duplicate data remaining. If DWS uses the upper function to do that, the source data will be changed, and it requires a lot of procedures and complex logics. To reduce the workload, the case-insensitive feature is added to DWS.

Collation Support

Collation Version Function

List partitioning

List partitioned tables are supported.

• Syntax:

  CREATE TABLE PARTITION

  ALTER TABLE PARTITION

• System catalog: PG_PARTITION
• Views:

  DBA_PART_INDEXES

  DBA_PART_TABLES

  DBA_TAB_PARTITIONS

  USER_IND_PARTITIONS

  USER_PART_INDEXES

  USER_TAB_PARTITIONS

Enhanced MySQL compatibility

ALTER TABLE can modify multiple columns at a time.

ALTER TABLE supports COMMENT.

ALTER TABLE supports CHANGE, DROP KEY, DROP PARTITION, and FIRST/AFTER.

ALTER INDEX, CREATE INDEX

ALTER TABLE, CREATE TABLE

ALTER TABLE PARTITION

Projection pushdown

Unnecessary columns in a multi-column view are automatically deleted during reference to avoid redundant calculation.

Added the description for projection pushdown to rewrite_rule

Share scan

Share scan

Stream plan CTEs support share scan.

Added enable_stream_ctescan.

The value STREAM_CTESCAN is added to logging_module.

The optional [NOT] MATERIALIZED is added to the WITH clause in the SELECT syntax.

Bloom Filter for column-store tables

In a HASH JOIN, if the thread of the foreign table contains HDFS tables or column-store tables, the Bloom filter is triggered.

enable_bloom_filter

Column-store replication tables support UPDATE.

DWS customers have a large number of MERGE INTO operations on replication tables. Like row-store replication tables, column-store replication tables now support UPDATE, DELETE, and MERGE INTO operations.

MERGE INTO

UPDATE

UPSERT

Fine-grained permission management

1. Added the ALTER, DROP and VACUUM permissions at table level.
2. Added the ALTER and DROP permissions at schema level.
3. Add the preset roles role_signal_backend and role_read_all_stats.

• Preset Roles
• Added table-level permissions:

  VACUUM

  VACUUM

  DROP TABLE

• Added schema-level permissions:

  ALTER SCHEMA

  DROP SCHEMA

  ALTER DEFAULT PRIVILEGES

• GRANT
• REVOKE

The problem of the INSERT OVERWRITE deadlock on multiple CNs is fixed.

No deadlock occurs when INSERT OVERWRITE is triggered on multiple CNs at the same time.

-

Execution plan optimization

• If the join conditions contain OR and the execution plan uses nestloop, the performance is poor. Therefore, the statement is changed to UNION ALL.
• To join column-store tables, if the join or filter conditions contain OR, the OR conditions are pruned before join.
• When large and small tables are joined across many DNs, the small table is broadcast instead of being redistributed.

Adjusting Key Parameters During SQL Tuning

Added projection_pushdown and or_conversion to rewrite_rule.

Relfile-based space statistics and space control

1. The query results of the pg_table_size, pg_relation_size and table_skewness views can be returned in seconds.
2. Schema space control can collect statistics on the full space and the constraints are removed.
3. The user/schema space calibration functions can be quickly executed, and the schema space calibration function can be automatically executed.
4. The SQL interface allocates space quota based on the total schema space instead of the DN level.

• Workspace Management
• Added GUC parameters for load management:

  space_once_adjust_num

  space_readjust_schedule

• Added the system catalog:

  PG_RELFILENODE_SIZE

• Added functions:

  gs_table_distribution(schemaname text, tablename text)

  gs_table_distribution()

  pgxc_wlm_readjust_relfilenode_size_table()

  pgxc_wlm_readjust_relfilenode_size_table(integer)

Enhanced performance fault locating and demarcation

1. Enhanced communication exception monitoring logs
2. Enhanced pgxc_stat_activity view
3. Top SQL statements can be associated with unique SQL statements.
4. Added the distributed view pg_session_wlmstat.

• Added the system view PG_COMM_QUERY_SPEED.
• Added the description for the lwtid and stmt_type column of PG_STAT_ACTIVITY.
• Added the unique_sql_id column information to GS_WLM_SESSION_HISTORY.
• Added the system view PGXC_SESSION_WLMSTAT.

Resource load management optimization and reconstruction

1. Enhanced user resource monitoring: The user resource monitoring logic is modified.
2. Enhanced resource monitoring: Job and user resource monitoring is optimized, and queue-level resource monitoring is added.

• User Resource Monitoring
• Enhanced resource monitoring:
  • Added: Resource Pool Monitoring.
  • Added system views:

    PGXC_RESPOOL_RESOURCE_INFO

    PGXC_RESPOOL_RUNTIME_INFO

    PGXC_RESPOOL_RESOURCE_HISTORY

    GS_RESPOOL_RESOURCE_INFO

    GS_RESPOOL_RUNTIME_INFO

  • Added the system catalog:

    GS_RESPOOL_RESOURCE_HISTORY

Modified partition table features

1. Changed the limit on the number of boundary values: The number of partitions in a partition table cannot exceed 32767, and the number of boundary values of all partitions cannot exceed 32767.
2. The default partition creation policy is modified. When a partition table is created, two default partitions are created. The partition time range of the two default partitions is PERIOD.

CREATE TABLE PARTITION

Redistribution column hint in the AGG process

During data migration from Oracle data warehouses, you can manually specify the distribution column using enhanced hints.

Plan Hint Optimization

Stream Operation Hints

Added the function of returning NULL when the divisor is 0. (Added in 8.1.3.110)

To specify whether to report an error when the divisor is 0 in MySQL compatibility mode, the compatibility configuration item enable_division_by_zero_mysql is added to the behavior_compat_options parameter.

behavior_compat_options

Added the DDL lock timeout configuration. (Added in 8.1.3.200)

Added the GUC parameter ddl_lock_timeout, which specifies the DDL lock timeout interval.

ddl_lock_timeout

Enhanced the view PG_LOCKS. (Added in 8.1.3.200)

• Added the views PGXC_WAIT_DETAIL and PGXC_LOCKWAIT_DETAIL.
• Add columns such as wait_on_pid and query_id.

• PGXC_WAIT_DETAIL
• PGXC_LOCKWAIT_DETAIL

Data masking compatibility configuration (Added in 8.1.3.310)

The GUC parameter redact_compat_options is added to configure the compatibility of the data masking techniques.

• redact_compat_options

Concurrent truncate, exchange, and select operations (added in 8.1.3.320)

Added the GUC parameter ddl_select_concurrent_mode. This parameter is used to solve the problem that the query statement lasts for a long time and blocks DDL in the scenario where the data volume is large or the query is complex. The effect is the same as that of the Oracle database.

Application scenarios:

• Concurrent truncate and select operations
• Concurrent exchange and select operations

Scenario constraints:

• Concurrency is not supported when there are conflicts with locks of higher levels (higher than level 1). For example, autoanalyze is triggered by select when autoanalyze_mode is set to normal.
• Concurrency is not supported when there are conflicts with locks in transaction blocks.

Connection constraints:

• When the JDBC, ODBC, or Python driver is used, if autocommit is set to false, this feature is not supported.
• This feature is not supported by query statements delivered by Data Studio and auto-committed transactions.

ddl_select_concurrent_mode

Enabling of the early stop optimization for Limit statements (added in 8.1.3.320)

Added the GUC parameter enable_limit_stop to specify whether to enable the early stop optimization for LIMIT statements.

enable_limit_stop

Early stop behavior change (Added in 8.1.3.322)

The default value of the GUC parameter enable_limit_stop is changed to on.

enable_limit_stop

Added

1

CREATE FOREIGN TABLE (SQL on Hadoop or OBS)

Added json for the parameter Format of OPTION.

2

CREATE FOREIGN TABLE (SQL on Hadoop or OBS)

Added force_mapping for OPTION.

3

Column type kvtype

Specifies the column type of time series tables in the CREATE TABLE and ALTER COLUMN syntax.

4

Table-level parameter: sub_partition_count

Configures the number of level-2 partitions in a time sequence table.

5

Table-level parameter ttl

Specifies the expiration time of auto-added partitions.

6

Table-level parameter period

Specifies the interval for triggering partition adding and dropping tasks and the time range of a single partition.

7

List partitioning support

The list partition operation syntax is added, including the CREATE and ALTER operations. The CREATE syntax is a separate syntax branch. In the AT syntax, except for the SPLIT syntax, other syntax are similar to those of the range partitioning.

8

ALTER TABLE ADD INDEX/DROP INDEX

The ALTER TABLE syntax branch is added to add and delete indexes.

9

CREATE TABLE ... LIKE

The CREATE TABLE... LIKE syntax without brackets is added.

10

ALTER TABLE ... ADD/MODIFY COLUMN

The statements for adding and modifying columns are added to ALTER TABLE. COMMENT and DEFAULT can be set.

11

Added the [NOT] MATERIALIZED syntax after with cte as.

Added the [NOT] MATERIALIZED syntax after with cte as.

12

alter index ... comment

The COMMENT clause is now available in the ALTER INDEX syntax.

13

ALTER TABLE ... add index comment

The COMMENT clause is now available in the alter table add index syntax.

14

create index index_name on table_name comment

The COMMENT clause is now available in the CREATE INDEX syntax.

15

copy to

Added server, bom, fileprefix, and maxrow to the OPTION of the COPY TO statement.

16

CREATE FOREIGN TABLE (OBS import and export)

Added bom to the OPTION of the CREATE FOREIGN TABLE statement (OBS import and export).

17

Operator =

If two Roaring Bitmaps are equal, true is returned. Otherwise, false is returned.

18

Operator <>

If two Roaring Bitmaps are not equal, true is returned. Otherwise, false is returned.

19

Operator &

Calculates the intersection of two Roaring Bitmaps.

20

Operator |

Calculates the union of two Roaring Bitmaps.

21

Operator |

Calculates the result of adding an ID to a Roaring Bitmap.

22

Operator #

Calculates the XOR result of two Roaring Bitmaps.

23

Operator -

Obtains the set of elements that are in a Roaring Bitmap but not in another Roaring Bitmap.

24

Operator -

Removes a specified ID from a Roaring Bitmap.

25

Operator @>

If the Roaring Bitmap before the operator contains the elements after the operator, true is returned. Otherwise, false is returned.

26

Operator <@

If the Roaring Bitmap after the operator contains the elements before the operator, true is returned. Otherwise, false is returned.

27

Operator &&

If two Roaring Bitmaps overlap, true is returned. Otherwise, false is returned.

28

ALTER TABLE DROP KEY

Added the syntax for deleting indexes. The usage of this syntax is the same as that of ALTER TABLE DROP INDEX.

29

ALTER TABLE CHANGE

Compatible with the CHANGE syntax in MySQL. Column names and column attributes can be modified at the same time.

30

ALTER TABLE ... FIRST/AFTER colname

The syntax layer supports the FIRST/AFTER syntax, but does not implement the actual semantics. It is applicable to the ALTER TABLE ADD/MODIFY/CHANGE COLUMN scenario and is controlled by the parameter skip_first_after_mysql.

31

AGG HINT

Added Agg hints for performance optimization.

32

ALTER TABLE ... ALTER COLUMN cstore_cu_sample_ratio

To reduce the work of CU decompression, samples are concentrated in randomly selected CUs. Therefore, it is difficult to obtain data features.

This field attribute is used to select more CUs without increasing the sampling ratio to facilitate data feature obtaining.

Modified

33

Table-level parameter deltarow_threshold

The default value of deltarow_threshold is set to 10000, which specifies the upper limit of rows when column-store tables are imported to the Delta table. This parameter is valid only when the table-level parameter enable_delta is enabled.

34

More data types are supported by Btree indexes.

The data types supported by the B-tree indexes are extended to support row store tables, column-store tables, and time series tables. btree_gin provides data types int2, int4, int8, float4, float8, timestamp with time zone, timestamp without time zone, time with time zone, time without time zone, date, interval, oid, money, "char", varchar, text, bytea, bit, varbit, macaddr, macaddr8, inet, cidr, uuid, name, bool, bpchar, and all enum types.

35

GRANT/REVOKE/ALTER DEFAULT PRIVILEGES

The ALTER/DROP/VACUUM permissions are added.

36

ALTER TABLE ... DROP PARTITION

The DROP PARTITION syntax can be used to delete multiple partitions.

37

DEFAULT clause in the CREATE/ALTER TABLE statement

The DEFAULT clause in the CREATE/ALTER TABLE statement does not support suffix operators (currently, only the factorial suffix operator "!" is used). For example, create table t (a int default 3!) is not allowed.

Added

1

MATERIALIZED

Added the [NOT] MATERIALIZED syntax after with cte as. As a non-reserved keyword, it can be still used as an object name. When it is used as a column alias, AS must be added.

2

time_fill

It is used as the keyword of a time filling expression to output the time_fill column. It cannot be used as a function name or user-defined data type name.

3

fill_first/fill_last/fill_avg

It is used as the keyword of a time filling expression to output the filled column. It cannot be used as a function name or user-defined data type name.

4

list

Specifies the type of a partitioned table. As a non-reserved keyword, it can still be used as an object name. When it is used as a column alias, AS must be added.

5

tsfield/tstag/tstime

Specifies the type of a partitioned table. As a non-reserved keyword, it can still be used as an object name. When it is used as a column alias, AS must be added.

Added

1

rb_added

A value is added to RoaringBitmap.

2

pg_partition

Added the boundexprs column to pg_partition.

3

pg_relfilenode_size

New system catalog

4

pg_attribute

The attkvtype column is added to pg_attribute to record the kvtype of columns.

5

pg_collation

The case_insensitive record is added to support case-insensitive behaviors.

Added

1

rb_build

Converts an int array into a bitmap.

2

rb_to_array

Reverse operation of rb_build. It converts a Roaring Bitmap into an int array.

3

rb_and

Obtains the intersection of two Roaring Bitmaps.

4

rb_or

Obtains the union of two Roaring Bitmaps.

5

rb_xor

Obtains the XOR result of two Roaring Bitmaps.

6

rb_andnot

Performs AND then negation operation on two Roaring Bitmaps.

7

rb_cardinality

Calculates the cardinality of a Roaring Bitmap.

8

rb_and_cardinality

Calculates the cardinality of the And result of two Roaring Bitmaps.

9

rb_or_cardinality

Calculates the cardinality of the OR result of two Roaring Bitmaps.

10

rb_xor_cardinality

Calculates the cardinality of the XOR result of two Roaring Bitmaps.

11

rb_andnot_cardinality

Calculates the cardinality of the andNot result of two Roaring Bitmaps.

12

rb_is_empty

Determines whether a Roaring Bitmap is empty.

13

rb_equals

Determines whether two Roaring Bitmaps are equal.

14

rb_intersect

Determines whether two Roaring Bitmaps intersect.

15

rb_contain

Determines whether the first Roaring Bitmap contains a specified value.

16

rb_add

Adds a value to a Roaring Bitmap.

17

rb_remove

Deletes a value from a Roaring Bitmap.

18

rb_flip

Reverses the Roaring Bitmap within a specified range.

19

rb_min

Obtains the minimum value of a Roaring Bitmap.

20

rb_max

Obtains the maximum value of a Roaring Bitmap.

21

rb_rank

Returns the cardinality of values in a bitmap that are less than or equal to the specified Offset.

22

rb_contain_rb

Check whether the first Roaring Bitmap contains the second Roaring Bitmap.

23

rb_containedby_rb

Check whether the second Roaring Bitmap contains the first Roaring Bitmap.

24

rb_containedby

Determines whether a specified value is included in a specified Roaring Bitmap.

25

rb_iterate

Returns the ints corresponding to a Roaring Bitmap.

26

rb_and_agg

Aggregates a Roaring Bitmap column based on the AND logic.

27

rb_or_agg

Aggregates a Roaring Bitmap column based on the OR logic.

28

rb_xor_agg

Aggregates a Roaring Bitmap column based on the XOR logic.

29

rb_and_cardinality_agg

Obtains the cardinality of the AND result of a Roaring Bitmap column.

30

rb_or_cardinality_agg

Obtains the cardinality of the OR result of a Roaring Bitmap column.

31

rb_xor_cardinality_agg

Obtains the cardinality of the XOR result of a Roaring Bitmap column.

32

rb_build_agg

Aggregates the int column into RoaringBitmap data.

33

pgxc_wlm_readjust_relfilenode_size_table()

Space statistics calibration function. It does not recreate the PG_RELFILENODE_SIZE system catalog but re-calibrates the user and schema space.

34

gs_table_distribution()

Quickly queries a table size.

35

pg_obs_cold_refresh_time

Modifies the time for automatic switchover of OBS multi-temperature tables.

36

gs_clean_tag_relation

Deletes the useless data in the tagid row of the tag table. The input parameter of the function is the OID of the time series table. Each partition traverses the minimum value of the tagid column in the cudesc table to obtain the minimum tagid value in the entire time series table.

37

proc_drop_partition

Drops a partition whose boundary time exceeds the TTL.

38

proc_add_partition

Creates partitions for a partitioned table.

39

pg_collation_actual_version

Returns the actual version number of an ICU collation.

40

first

Indicates the first element in a group.

41

last

Indicates the last element in a group.

42

mode

Indicates the value with the highest occurrence frequency in a group.

43

delta

Indicates the difference between two adjacent rows.

44

percentile_of_value

Indicates the approximate percentile value.

45

value_of_percentile

Indicates the approximate percentile.

46

spread

Indicates the difference between the maximum value and minimum value in a group.

47

pg_flush_buffers

Flushes all row-store dirty pages.

Modified

48

pg_stat_activity series views

The stmt_type and lwtid columns are added to the pg_stat_get_activity_with_conninfo, pg_stat_activity, pgxc_stat_activity and pg_stat_get_activity views.

49

Added two rows to the pg_authid system catalog.

Added the preset roles pg_role_signal_backend and pg_role_read_all_stats.

50

vac_fileclear_relation

Returns the number of files to be cleared in a specified table. If no files are to be cleared, 0 is returned.

51

vac_fileclear_all_relation

Returns the number of files to be cleared in all column store tables. If no files are to be cleared, 0 is returned.

Added

1

pgxc_session_wlmstat

New view

2

pg_comm_query_speed

New view. It is used to query sending information based on query_ids.

3

pgxc_respool_resource_info

pgxc_respool_runtime_info

pgxc_respool_resource_history

gs_respool_resource_info

gs_respool_runtime_info

New views related to resource monitoring.

4

pgxc_wait_detail

pgxc_lockwait_detail

Added the enhanced view pg_locks.

Modified

5

DBA_PART_INDEXES

DBA_PART_TABLES

DBA_TAB_PARTITIONS

USER_IND_PARTITIONS

USER_PART_INDEXES

USER_TAB_PARTITIONS

Added description for list partitions.

6

pgxc_wlm_session_statistics

The logic for querying real-time information about Top SQL statements from all CNs is changed from serial to parallel to improve performance. The function remains unchanged.

7

all_indexes

The definition of the view all_indexes in sys and pg_catalog is incorrect. If objects with the same name exist in different schemas, the result set expands.

Modified

1

In CREATE INDEX, the target table is a time series table.

Any index created in the time series table is converted to a double index in the tag table. The index column of the double index is the specified column that the index is created for.

2

Secondary load management

Added secondary management to load management. Functions and multi-statements that contain complex queries may trigger multiple controls. You can set enable_transaction_parctl to off to disable secondary management. If you do so, transaction block statements and multi-statement management will also be disabled.

3

Load management (autoanalyze management)

The control logic for queries triggering autoanalyze is changed from no control to control. You can set enable_transaction_parctl=off to disable the autoanalyze control.

4

User monitoring view pg_total_user_resource_info

1. Changed the CPU and memory resource usage and limit to the resource usage and limit in the cluster.
2. The monitored CPU, I/O, and memory objects are changed from complex jobs to all jobs.
3. The CPU monitoring logic is changed from Cgroup monitoring to job CPU monitoring summary.

5

Audit logs

1. Statements in a transaction are audited when the statement type is set to be audited, even if the transaction is not set to be audited.

2. The DECLARE CURSOR statement is audited when the guc parameter audit_operation_exec is set to select.

Added

1

space_readjust_schedule

Indicates whether to enable automatic calibration for users and schema spaces. The options are off, auto, and auto (xxx K/M/G). The default value is auto.

2

space_once_adjust_num

Threshold of the number of files processed each time during slow build and fine-grained calibration in the space management and statistics functions. The default value is 300.

3

tag_cache_max_number

Indicates the maximum number of items cached in the tag cache of a global hash table. The default value is 10000000.

4

enable_tagbucket_auto_adapt

Indicates whether to enable tagbucket adaptive adjustment. The default value is on.

5

cache_tag_value_num

Number of cached tag tuples during tag column lateread. The default value is 60000.

6

buffer_ring_ratio

Controls the threshold and switch of the ring buffer in a row-store table. The default value is 250, that is, 1/4 (250/1000), which is the same as the previous logic.

7

enable_stream_ctescan

This parameter is added. By default, this parameter is set to OFF after an upgrade and is set to ON after a new installation. Specifies whether to enable the share scan function for the stream plan.

8

behavior_compat_options

The disable_including_all_mysql option is added to control whether the CREATE TABLE LIKE syntax is in INCLUDING ALL mode. By default, this option is not set and the CREATE TABLE LIKE syntax is in INCLUDING ALL mode.

9

profile_logging_module

Added the GUC parameter profile_logging_module to configure the type of performance logs. By default, OBS, Hadoop, and REMOTE_DATANODE are enabled and MD is disabled after upgrade or installation. The method of setting this parameter is the same as that of setting logging_module.

10

object_mtime_record_mode

Controls the behaviors of pg_object recording modification time. By default, the behavior of the old version is retained. The new option can be set to not record the TRUNCATE/GRANT/REVOKE operations on partitions.

11

skew_option

Specifies whether to enable an optimization policy.

12

ddl_lock_timeout

DDL lock timeout interval.

Modified

13

rewrite_rule

• The orconversion option is added and is enabled by default.

  Indicates the OR optimization rules for equi-join. Change: The query plan converts nestloop to hashjoin.

• The projection_pushdown option is added and is enabled by default.

  Change: Unused columns in subqueries, CTEs, and views are deleted.

14

default_orientation

The userset guc parameter default_orientation is added to control the default type of tables users can create (orientation is not set). The options are as follows:

Row-store table (row), column-store table (column), and column-store table with delta enabled (column enabledelta)

15

max_process_memory

It is an OM setting item configured during initial installation. The original formula is: Available memory x 0.7/(1 + Number of DNs). The coefficient 0.7 is changed to 0.8. It remains unchanged in scenarios such as upgrade and scale-out of the old cluster.

16

enable_bloom_filter

Reused parameter. It is now used to control BloomFilter in a column-store table. The default value remains unchanged (Enabled).

17

retry_ecode_list

Added the default value 45003 to solve the problem that wrong partitions may be queried when partitions are added and queried concurrently in a list partitioned table.

18

auth_iteration_count

The default value is reduced from 50000 to 10000. The change applies to both installation and upgrade. Indicates the number of hash iterations of the passwords of the client and server.