2024

Category

Feature

Description

References

New feature

Storage engine

The data cache feature is supported for the Beam engine to accelerate nested-loop index joins.

Beam storage engine

Automatic partitioning

By default, the automatic partitioning feature is enabled for partitioned tables.

Automatic partitioning in AnalyticDB for PostgreSQL V7.0

Partition dump

The dump_table_ddl() function can be used to export the schema of partitioned tables.

None

Optimized feature

Planner

The LIMIT query plans in the planner are optimized to improve the performance of index scans and LIMIT queries.

None

Optimizer

The performance of the static partition pruning feature of the Orca optimizer is improved.

None

Fixed issue

Optimizer

The following issue is fixed: The Orca optimizer fails to generate execution plans that contain RIGHT JOIN for common table expression (CTE) queries.

None

Partitioned table

The following issue is fixed: When the ANALYZE statement is executed on a partitioned table, the Not a Number (NaN) value is returned for the correlation parameter in specific scenarios. This may result in abnormal index scan costs.

None

Storage engine

• The following issue is fixed: When the auto-vacuum operation is performed on a Beam table, data compaction for which high-level locks are configured is unexpectedly triggered.

• The following issue is fixed: When Data Transmission Service (DTS) is used to synchronize the data of a Beam table, data flushing fails to be triggered.

• The following issue is fixed: Speculative insertion of Beam data fails to trigger callback functions.

Beam storage engine

Stability

• Disk scaling is enhanced to prevent exhausted disk space on the coordinator node.

• Dictionary encoding issues are fixed.

• The following issue is fixed: An instance fails to be restarted because the number of connections reaches the upper limit.

Dictionary encoding

Storage engine

The following issue is fixed: The cache performance is degraded due to the reading of residual records.

None

Category

Feature

Description

References

Optimized feature

Planner

The LIMIT query plans in the planner are optimized to improve the performance of index scans and LIMIT queries.

None

Fixed issue

Optimizer

The following issue is fixed: The Orca optimizer fails to generate execution plans that contain RIGHT JOIN for CTE queries.

None

Stability

• The following issue is fixed: An instance fails to be restarted because the number of connections reaches the upper limit.

• The following issue is fixed: When the ANALYZE statement is executed on a partitioned table, the NaN value may be returned for the correlation parameter in specific scenarios. This may result in abnormal index scan costs.

• Disk scaling is enhanced to prevent exhausted disk space on the coordinator node.

None

Storage engine

• The following issue is fixed: When the auto-vacuum operation is performed on a Beam table, data compaction for which high-level locks are configured is unexpectedly triggered.

• The following issue is fixed: When DTS is used to synchronize the data of a Beam table, data flushing fails to be triggered.

• The following issue is fixed: Speculative insertion of Beam data fails to trigger callback functions.

Beam storage engine

Category

Feature

Description

References

New feature

Vector search

The vector search engine is fully supported in AnalyticDB for PostgreSQL V7.0.

None

Synonym management

The synonym search feature is supported.

Use synsets

PostgresML

An end-to-end platform that integrates AI/ML and Retrieval-Augmented Generation (RAG) is supported based on the AI/ML and RAG features of PostgreSQL.

In-database AI/ML

Asynchronous materialized view

Asynchronous materialized views are fully supported in AnalyticDB for PostgreSQL V7.0. When data in a base table changes, the materialized view created based on the base table can be automatically updated.

None

Optimized feature

Executor

The performance of the Laser engine at the execution layer is improved.

None

Optimizer

The performance of the static partition pruning feature of the Orca optimizer is improved.

None

Fixed issue

Stability

Issues related to the analyze and vacuum operations and sort keys of the Beam storage engine are fixed.

None

Dictionary encoding

Dictionary encoding issues are fixed.

Dictionary encoding

Category

Feature

Description

References

Optimized feature

Data link

The MaxCompute foreign data wrapper (FDW) feature is updated to improve stability.

None

Query acceleration

The query cache feature is updated to improve stability.

Query cache in AnalyticDB for PostgreSQL V7.0

Fixed issue

Stability

Issues related to the analyze and vacuum operations and sort keys of the Beam storage engine are fixed.

Use auto-vacuum

Beam usage

Category

Feature

Description

References

Fixed issue

Minor version update

The minor version update failure is fixed by modifying the installation logic of the minor version update extension.

Update the minor version of an instance

Scaling

The redo xlog failure that occurs when added nodes are started is fixed.

Scale an instance

Data access

The data access interruption issue caused by limits is fixed.

None

Category

Feature

Description

References

New feature

Multi-zone deployment

The multi-zone deployment feature is supported to implement disaster recovery in the same region. This ensures a recovery point objective (RPO) of 0.

None

High-performance full-text search

High-performance full-text search based on the Best Matching 25 (BM25) algorithm is supported.

High-performance full-text search (public preview)

FIXDECIMAL

The FIXDECIMAL type is supported to replace the DECIMAL or NUMERIC type to improve query speed.

None

High availability

The hot mirror feature is supported.

None

Optimized feature

Automatic partitioning

Partitions can be automatically created based on inserted values.

None

Beam storage engine

The performance of the bitmapscan feature is improved.

Beam storage engine

Optimizer

• Functions that involve distributed tables can be executed in distributed query statements.

• The optimizer_force_agg_skew_avoid parameter is added for the Orca optimizer. The parameter can be used to determine whether a constant is included in the GROUP BY column to prevent data skew.

• Risk control is optimized for the Broadcast and Gather motions.

• The two-stage window aggregation feature is supported for the Orca optimizer.

• Specific stability issues are fixed.

Optimizers

Fixed issue

Scaling

• The scaling time estimation error is fixed.

• The memory setting issue of the table-level parameter reduce_mode is fixed.

Scale an instance

Logical backup

The extension installation issue that occurs during an upgrade is fixed.

None

Storage engine

The stability issue, such as a crash occurring when data is written to an append-only or Beam table that has generalized inverted index (GIN) indexes, is fixed.

Storage engine

Optimizer

The execution plan failure in specific partitioned tables is fixed for the Orca optimizer.

Optimizers

Execution engine

Specific executor stability issues are fixed.

Execution engine

Category

Feature

Description

References

New feature

Query cache

Query results can be cached to accelerate the response to repeated queries.

Query cache in AnalyticDB for PostgreSQL V7.0

Optimized feature

Tiered storage

Automatic data migration from hot storage to cold storage is optimized to adapt to the Beam storage engine.

Automatic data migration from hot storage to cold storage

Minor version update

The shared system table logic can be modified.

Update the minor version of an instance

Dump

When the pg_dump tool is used to dump a partitioned table, all partitions of the table are dumped.

None

Stability

The logic that is used to extend global transaction ID (GXIDs) from 32 bits to 64 bits is optimized.

None

Data masking

The copy and equal functions can be used for data masking.

Dynamic data masking

Optimizer

• The window aggregation feature is supported for non-correlated queries.

• The execution plan of MaxCompute external tables is optimized to improve the performance of associated queries.

Optimizers

Fixed issue

Storage engine

• The following issue is fixed: The Beam storage engine ignores the trailing fragment if a large number of dead tuples exist when the analyze operation is performed.

• The following issue is fixed: The compaction resource limits of the Beam storage engine do not take effect.

• The following issue is fixed: A subtransaction that involves Delta and Base tables results in incorrect metadata.

Beam storage engine

Stability

The issue about residual 2-phase commit transactions in the PostgreSQL community is fixed.

None

Optimizer

• The stability issue of optimizers and operators is fixed.

• The following issue is fixed: An error related to unexpected hash join plans occurs when dictionary encoding is used to accelerate queries.

Optimizers

Real-time materialized view

The following issue is fixed: An error occurs on automatic nesting when an OUTER JOIN operation is performed on a real-time materialized view.

Real-time materialized views

Category

Feature

Description

References

New feature

pgsearch extension

The pgsearch extension is supported to perform full-text search in vector databases.

None

pg_buffercache extension

The pg_buffercache extension is supported to check whether a table is cached.

pg_buffercache

Vector algorithms in vector search

The following distance metrics that can be used together with operator symbols are supported for vector search: L1, Hamming, and Jaccard. The Hierarchical Navigable Small World (HNSW) algorithm is supported.

Perform vector search

Fine ranking in vector search

Fine ranking is supported for vector search.

Perform vector search

Optimized feature

Vector indexes can be created based only on memory without being written to disks in vector search.

More comprehensive logic is supported for modifying system tables if an AnalyticDB for PostgreSQL instance is updated to the latest version.

Subqueries can be pulled up by the optimizer for projection in queries.

A custom format for reading JSON data is supported.

Information about the table access method is included in the output DDL statements when the pg_dump tool is used to dump a table.

Features of resource groups are optimized.

Fixed issue

Data can be moved from tables that contain column names enclosed in quotation marks ('') to cold storage.

The performance and visibility issues of the Beam storage engine are fixed.

The issue that causes startup failures of secondary compute nodes when the gpstart tool is used is fixed.

The following issue is fixed: The iconv tool fails to convert the character set encoding format from GBK into UTF-8 in an Alinux3 environment.

The following issues are fixed: A scale-in operation times out. Data is lost when you perform concurrent rewrite and insert operations on partitioned tables.

The following issue is fixed: Core dumps occur when the Laser engine is executed.

Category

Feature

Description

References

Optimized feature

Update of GanosBase to version 6.7

AnalyticDB for PostgreSQL GanosBase is updated to version 6.7. GanosBase provides a series of data types, functions, and stored procedures for AnalyticDB for PostgreSQL to store, index, query, analyze, and compute spatio-temporal data in a cost-effective manner.

Overview

Fixed issue

The following issue is fixed: After you extend a partitioned table, data errors occur on the table because the values of specific fields are excessively long in length.

The following issue is fixed: Standard accounts do not have the permissions to use the auto_partition extension to configure automatic partition management.

The following issue is fixed: When you execute the ALTER TABLE RESET or ALTER TABLE SET statement by using the Beam storage engine, an error occurs on queries that involve table indexes created by using earlier versions of Beam.

The following issue is fixed: An error occurs when you attempt to add a column with a NOT NULL constraint to a Delta table by using the Beam storage engine.

The following issue is fixed: When you perform a query by using CTEs, the ERROR: could not find CTE error message is returned in the execution plan due to pull-up of an CTE.

The following issue is fixed: In scenarios that involve an external data source, core dumps occur when a single session sequentially accesses the two extensions, PostgreSQL FDW and Greenplum FDW.

Category

Feature

Description

References

New feature

Security feature

Row-level permission control on materialized views is supported.

None

Fixed issue

The following issue is fixed: Residual fragments are reclaimed by the vacuum fragment auxiliary table of a Beam table after an AnalyticDB for PostgreSQL instance crashes.

The following issue is fixed: WAL logs are lost when you use the pg_rewind tool to perform incremental data restoration.

The following issue is fixed: Database accounts that have the RDS_SUPERUSER permission cannot use the auto_partition extension.

Category

Feature

Description

References

New feature

Real-time materialized view

Real-time materialized views are supported. Compared with standard materialized views, real-time materialized views can be automatically refreshed when data changes without the need to execute REFRESH statements.

Real-time materialized views

Optimized feature

Support for writes during the scale-out of the Beam storage engine

Write operations, such as INSERT ON CONFLICT and COPY ON CONFLICT, can be performed on Beam tables during the scale-out of the Beam storage engine.

Overview

Fixed issue

The following issue is fixed: Core dumps occur when you execute the VACUUM AUX_ONLY or VACUUM AO_AUX_ONLY statements on a parent partitioned table.

The following issue is fixed: A database account that has the RDS_SUPERUSER permission cannot grant permissions to standard accounts.

The following issue is fixed: A mapping logic error occurs or core dumps occur when the CHAR type in the Optimized Row Columnar (ORC) format is mapped to the CHAR type in AnalyticDB for PostgreSQL in scenarios that involve Object Storage Service (OSS) foreign tables.

Category

Feature

Description

References

Fixed issue

The following issue is fixed: Data loss occurs due to incorrect subtransaction visibility logic when you concurrently write transactions.

The following issue is fixed: Blocked files fail to be cleared when the backup archiving process unexpectedly fails. This issue may block subsequent xlog archiving and recycling processes and cause accumulated local log files.

The following issue is fixed: The negative bitmapset member not allowed error message is returned when the storage engine processes specific data. This issue may cause unexpected execution engine failures.

The following issue is fixed: The beam could not read block in file error message is returned when the system attempts to read specific data file blocks. The file reading logic is optimized to enhance error tolerance for damaged or unreadable file blocks. This ensures stability and reliability when data is read.

The following issue is fixed: The monitoring query tool unexpectedly fails. The underlying implementation is optimized to improve error handling capabilities and stability and prevent failures.

Category

Feature

Description

References

Optimized feature

Beam

• The WHERE condition is supported when you execute the INSERT ON CONFLICT and COPY ON CONFLICT statements to update data.

• Partitioned tables are supported for the COPY ON CONFLICT statement.

None

Laser

Stored procedures, cursors, and the following statements are supported: INSERT, UPDATE, DELETE, CTAS, REFRESH MATERIALIZED VIEW, and COPY.

None

Orca and planner optimizers

• The B-tree index cost estimation logic of the adbpg_optimizer_costmodel parameter is optimized. By default, the optimizer feature is disabled.

• Materialized views can be updated for the Orca optimizer.

None

Slow queries

The CPU utilization, memory usage, and disk file sizes of each query can be viewed at the instance and node levels.

None

Fixed issue

The following issue is fixed: Core dumps occur when you execute the ANALYZE statement on append-optimized column-oriented (AOCO) tables.

The following issue is fixed: Data fails to be moved to cold storage when no partitioned table name is specified or the NOT NULL constraint is specified for the parent table.

The following issue is fixed: The variable not found in subplan targetlist error message is returned when an SQL statement that contains AGG Filter clauses is parsed.

The following issue is fixed: Core dumps occur when you add a sort key.

The following issue is fixed: Temporary files remain in specific scenarios.

The following issue is fixed: DDL operations are blocked during a scale-in activity.

The following issue is fixed: The Query Monitor extension unexpectedly stops working.

The following issue is fixed: Data fails to be read and written when you rebuild indexes during a scale-out activity.

Category

Feature

Description

References

Fixed issue

The following issue is fixed: The reloptions attribute of the parent partitioned table is incorrect after you add columns to Beam partitioned tables.

The following issue is fixed: Auxiliary tables are inconsistent after you execute an ALTER statement that contains multiple substatements on a Beam partitioned table.

Category

Feature

Description

References

Fixed issue

The following issue is fixed: An error occurs when you concurrently execute the INSERT ON CONFLICT and DROP TABLE statements.

The following issue is fixed: The Cancelled BY Global Deadlock Detector error message is returned when you perform concurrent writes to Beam tables.

The pg_relation_size compatibility issue in earlier versions is fixed.

The following issue is fixed: Core dumps occur due to incorrect sort keys when you scale out materialized views.

The following issue is fixed: Core dumps occur in V7.0.5.0 when vector data fails to be constructed.

Category

Feature

Description

References

New feature

Elastic scaling

The elastic scaling feature for adding compute nodes is supported. When you add compute nodes, you can read, write, and update tables without business interruption.

Change compute node configurations

Fixed issue

The following issue is fixed: Incorrect execution plans are generated by the Orca optimizer when the JOIN ON clause contains cross-level subqueries.

The following issue is fixed: Core dumps occur when you use the Orca optimizer to perform the UNION ALL and SORT operations.

The following issue is fixed: An error occurs when you execute the COPY ON CONFLICT DO NOTHING statement due to data conflicts.

The following issue is fixed: Beam flush operations are frequently canceled, which may cause performance degradation.

The dictionary memory issue of the Beam storage engine is fixed.

The following issue is fixed: The Query Monitor extension unexpectedly stops working.

The following issue is fixed: System tables may contain residual data when you execute the DROP FOREIGN SERVER statement on foreign tables.

The security issue is fixed. Security hardening is performed.

The compatibility issue that occurs when you execute the Laser Scan operator on ctid columns is fixed.