2025

Description

Related documentation

This version is compatible with community PostgreSQL 16.10.

Community PostgreSQL 16.10

Added a feature that allows privileged users to load extensions by using the LOAD command.

-

The oracle_fdw, pg_partman, pg_bitemporal, and plv8 extensions are now supported. In addition, the pg_duckdb extension is upgraded to version 1.0.

-

Description

Optimized the global execution plan cache (GPC) mechanism. The GPC no longer caches query plans with custom scans.

Optimized the data synchronization mechanism for In-Memory Column Index (IMCI) on read-only (RO) nodes to improve data freshness in the columnstore index.

Optimized the data rebalance operation for a distributed cluster by adding parallel processing capabilities.

Optimized the execution performance of the pg_total_relation_size function.

Description

Fixed a syslogChannel initialization exception.

Fixed a memory leak in the SMgrRelation object that could occur in specific scenarios.

Fixed an issue where some backend worker processes were not registered in the pg_stat_activity view.

Fixed an issue that could cause slab memory to continuously increase during the replay of two-phase commit (2PC) transaction logs.

Description

Related documentation

Added support for Commit Sequence Number (CSN) to improve Multi-Version Concurrency Control (MVCC) performance.

Global Consistency

Added support for the OUTLINE feature. This feature automatically adds specified hints to certain types of SQL statements. This allows precise control over SQL execution plans without affecting business applications.

Plan fixation (STATEMENT OUTLINE)

Audit logs can now include bound SQL parameters from the extended protocol, providing more complete records for troubleshooting and security audits.

-

A distributed cluster now supports static pruning for IN clauses. This feature filters out irrelevant data shards before query execution, which reduces the amount of data scanned and improves distributed query performance.

-

A distributed cluster now supports parallel data rebalance between colocation groups. This significantly improves data rebalancing efficiency and reduces the execution time of DDL operations.

-

Added support for the following extensions: credcheck, decoderbufs, ip4r, pg_permissions, pg_profile, pg_stat_kcache, pgaudit, polar_sql_inception SQL audit extension, postgresql_anonymizer, sequential_uuids, and wal2json.

• age: Added the agtype_increase function to perform accumulation operations on attribute data.

• DocumentDB: Updated to version 0.105-0, which adds support for the listDatabase command and displays the sizeOnDisk value.

• FerretDB is updated to version 2.4.0 and supports the buildInfo and lastError commands in OP_Query.

• pgvector: Supports bulk write.

• polar_ai: Added chatbi and model fine-tuning functions.

• decoderbufs (decoding to ProtoBuf)

• ip4r (IP address and network storage)

• pg_profile (resource monitoring)

• pgaudit

• wal2json

Description

Optimized the asynchronous lock replay logic on secondary databases. This improves lock operation efficiency during physical replication and reduces query latency on read-only nodes.

Improved the data synchronization efficiency between row stores and In-Memory Column Index (IMCI). This accelerates the synchronization of data from row stores to columnstore indexes and ensures that analytical queries can run on more up-to-date data.

Improved cluster startup performance and reduced resource usage. The pldebugger (stored procedure debugger) extension is no longer loaded by default.

Description

Fixed an issue where an improper readahead history reset condition could decrease the cache hit ratio or cause I/O performance fluctuations.

Fixed an issue where a Serverless cluster could crash during scale-out due to the Parallel Buffer flushing an incorrect cache ID.

Fixed an issue in the console where some modifiable parameters were incorrectly displayed as non-modifiable.

Fixed an issue where the ai_buildSchemaIndex function in the polar_ai extension could generate an incorrect table structure when processing pattern-type indexes.

Description

Related documentation

Added support for read-only nodes for columnstore indexes.

Enable and use columnstore indexes

Added support for the extensions pg_bitemporal, polar_age, hypopg, imgsmlr, and pg_partman.

• polar_ai: Supports the NL2SQL feature.

• polar_age:

  • The label table automatically adds a timeline field to support the graph data Time to Live (TTL) feature.

  • You can manually add a vector field to the label table to enable integrated storage and querying of graphs and vectors for use cases such as the long-term memory feature.

• Graph Database

• hypopg (hypothetical indexes)

• imgsmlr (image similarity)

• Graph analysis based on PolarDB: A one-stop solution for AI agent long-term memory

polar_monitor_preload is added to the preload libraries by default.

-

Description

Optimize pg_total_relation_size performance.

Optimize CREATE TABLE AS performance.

Description

Related documentation

This version is compatible with community PostgreSQL 16.9.

Community PostgreSQL 16.9

Supports distributed HINT, and is used in the same way as pg_hint_plan.

pg_hint_plan

EXPLAIN statements support the use of HINTs for the polar_csi.enable_query parameter of the vectorized engine.

Instructions for using vectorized engine parameters

Added support for the pg_hint_plan, pg_sphere, smlar, DocumentDB, pg_graphql, pgjwt, pgnet, and vault extensions.

• pg_hint_plan

• pg_sphere (spherical data calculation)

• smlar (similar vectors)

Description

Improve the execution speed of pg_switch_wal.

Optimized the data synchronization speed for large wide tables in the vectorized engine.

Description

Fixed a memory leak in the on-demand replay of LogIndex.

Fixed an issue where polar_stat_proxy was not initialized.

Fixed an issue where the replay logic for XLOG_BTREE_REUSE_PAGE in parallel replay did not use a 64-bit FullTransactionId, which resulted in repeated replay conflict checks.

Description

Related documentation

This version is compatible with community PostgreSQL 16.8.

Community PostgreSQL 16.8

Added support for the vectorized engine to better handle complex queries.

Vectorized engine

The spatio-temporal database (Ganos) is upgraded to version 7.4.

Spatio-temporal database release notes

Added support for the age and rum extensions.

• age

• rum

Description

Supports the multi-insert feature, enabling batch writes.

Improved the promotion speed of standby nodes.

Description

Related documentation

You can use channel to scan external data over a secure network channel.

CREATE SERVER

The spatio-temporal database (Ganos) is upgraded to version 7.2.

Spatio-temporal database release notes

Added support for force-flushing hot pages by using full-page writes.

-

Added support for the mysql_fdw, oss_fdw, and wal2json extensions.

• mysql_fdw

• oss_fdw

• wal2json

Description

Added support for Read Stream and merged read I/O to accelerate data queries.

Added support for batch reads of Write-Ahead Logging (WAL) logs during the recovery phase.

Optimized the index creation process.

Added support for batch extension of table files to accelerate data import and crash recovery.