Amazon Aurora PostgreSQL updates - Amazon Aurora
PostgreSQL 14.5PostgreSQL 14.4PostgreSQL 14.3PostgreSQL 13.8PostgreSQL 13.7PostgreSQL 13.6PostgreSQL 13.5PostgreSQL 13.4PostgreSQL 13.3PostgreSQL 12.12PostgreSQL 12.11PostgreSQL 12.10PostgreSQL 12.9PostgreSQL 12.8PostgreSQL 12.7, Aurora 4.2PostgreSQL 12.6, Aurora 4.1 (Deprecated on July 15, 2022)PostgreSQL 12.4, Aurora 4.0 (Deprecated on July 15, 2022)PostgreSQL 11.17PostgreSQL 11.16PostgreSQL 11.15PostgreSQL 11.14PostgreSQL 11.13PostgreSQL 11.12, Aurora 3.6PostgreSQL 11.11, Aurora 3.5 (Deprecated on July 15, 2022)PostgreSQL 11.9, Aurora 3.4PostgreSQL 11.8, Aurora 3.3 (Deprecated on July 15, 2022)PostgreSQL 11.7, Aurora 3.2 (Deprecated on March 15, 2022)PostgreSQL 11.6, Aurora 3.1 (Deprecated on March 15, 2022)PostgreSQL 11.4, Aurora 3.0 (Deprecated on October 28, 2021)PostgreSQL 10.21 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)PostgreSQL 10.20 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)PostgreSQL 10.19 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)PostgreSQL 10.18 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)PostgreSQL 10.17, Aurora 2.9 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)PostgreSQL 10.16, Aurora 2.8 (Deprecated on July 15, 2022)PostgreSQL 10.14, Aurora 2.7 (Deprecated on July 15, 2022)PostgreSQL 10.13, Aurora 2.6 (Deprecated on July 15, 2022)PostgreSQL 10.12, Aurora 2.5 (Deprecated on March 15, 2022)PostgreSQL 10.11, Aurora 2.4 (Deprecated on March 15, 2022)PostgreSQL 10.7, Aurora 2.3 (Deprecated on October 28, 2021)PostgreSQL 10.6, Aurora 2.2 (Deprecated on October 28, 2021)PostgreSQL 10.5, Aurora 2.1 (Deprecated on October 28, 2021)PostgreSQL 10.4, Aurora 2.0 (Deprecated on October 28, 2021)PostgreSQL 9.6.22, Aurora 1.11 (Deprecated on January 31, 2022)PostgreSQL 9.6.21, Aurora 1.10 (Deprecated on January 31, 2022)PostgreSQL 9.6.19, Aurora 1.9 (Deprecated on January 31, 2022)PostgreSQL 9.6.18, Aurora 1.8 (Deprecated on January 31, 2022)PostgreSQL 9.6.17, Aurora 1.7 (Deprecated on January 31, 2022)PostgreSQL 9.6.16, Aurora 1.6 (Deprecated on January 31, 2022)PostgreSQL 9.6.12, Aurora 1.5 (Deprecated on October 28, 2021)PostgreSQL 9.6.11, Aurora 1.4 (Deprecated on October 28, 2021)PostgreSQL 9.6.9, Aurora 1.3 (Deprecated on October 28, 2021)PostgreSQL 9.6.8, Aurora 1.2 (Deprecated on October 28, 2021)PostgreSQL 9.6.6, Aurora 1.1 (Deprecated on October 28, 2021)PostgreSQL 9.6.3, Aurora 1.0 (Deprecated on October 28, 2021)
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China.

Amazon Aurora PostgreSQL updates

Following, you can find information about versions of the Amazon Aurora PostgreSQL-Compatible Edition database engine that have been released for Amazon Aurora. Many of the listed releases include both a PostgreSQL version number and an Amazon Aurora version number. However, starting with the release of PostgreSQL versions 13.3, 12.8, 11.13, 10.18, and for all other later versions, Aurora version numbers aren't used. To determine the version numbers of your Aurora PostgreSQL database, see Identifying versions of Amazon Aurora PostgreSQL in the Amazon Aurora User Guide.

For information about extensions and modules, see Extension versions for Amazon Aurora PostgreSQL.

Amazon Aurora PostgreSQL 1.X (compatible with PostgreSQL 9.6.XX) reached end of support on January 31, 2022. Upgrade your databases that are running Aurora PostgreSQL 9.6 to Aurora PostgreSQL 11 or higher now. To learn how, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

If you are running Amazon Aurora PostgreSQL 2.X (compatible with PostgreSQL 10.XX), make sure to upgrade your databases to Amazon Aurora PostgreSQL 11 or higher before January 31, 2023. To learn how, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide. For more information, see Announcement: Amazon Aurora PostgreSQL 10.XX end of support is January 31, 2023..

If you are running Amazon Aurora PostgreSQL minor versions 10.13, 10.14, 10.16, 11.8, 11.11, 12.4, or 12.6, make sure to upgrade your databases before July 15, 2022:

  • For PostgreSQL minor versions 10.13, 10.14 and 10.16, the recommended minimum minor version is 10.17.

  • For PostgreSQL minor versions 11.8 and 11.11, the recommended minimum minor version is 11.12.

  • For PostgreSQL minor versions 12.4 and 12.6, the recommended minimum minor version is 12.7.

To learn how, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

For more information about Amazon Aurora available releases, policies, and time lines, see How long Amazon Aurora major versions remain available in the Amazon Aurora User Guide. For more information about support and other policies for Amazon Aurora see Amazon RDS FAQs.

To determine which Aurora PostgreSQL DB engine versions are available in an Amazon Web Services Region, use the describe-db-engine-versions Amazon CLI command as shown following.

aws rds describe-db-engine-versions --engine aurora-postgresql --query '*[].[EngineVersion]' --output text --region aws-region

For a list of Amazon Web Services Regions, see Aurora PostgreSQL Region availability in the Amazon Aurora User Guide.

Topics

PostgreSQL 14.5

This release of Aurora PostgreSQL is compatible with PostgreSQL 14.5. For more information about the improvements in PostgreSQL 14.5, see PostgreSQL release 14.5.

Aurora PostgreSQL 14.5.0, November 09, 2022

High priority stability enhancements

  • Fixed an Aurora Serverless v2 scaling issue.

  • Fixed an issue which caused Aurora Serverless v2 shrink failure.

General stability enhancements

  • This release contains all of the fixes, features, and improvements present in PostgreSQL release 14.4.

  • Improved buffer cache scavenging when the buffer cache is in duress.

  • Fixed an issue in Database Activity Streams that leads to high memory consumption.

  • Fixed an issue where DB instances could restart when applying changes from a logical replication subscription.

  • Fixed an issue that caused DB instance restarts.

  • Fixed an issue where a DB instance restarts recursively while generating monitoring metrics during a crash.

  • Fixed an issue where a DB instance restarted during performance metric collection.

  • Fixed an issue where an attempt to connect to the database would fail with SSLV3_ALERT_CERTIFICATE_UNKNOWN.

  • Improved the error reporting in case of an inconsistent B-tree index.

  • Improved the diagnostic logging around setting invalid hint bits.

  • Fixed an issue where autovacuum would incorrectly skip tables.

  • Improved the logical replication prefetching.

  • Fixed a durability issue in the GIN indexes.

  • Provided options to configure MultiXact SLRU cache. By setting the following parameters (GUCs), customers will now be able to change the MultiXact SLRU cache sizes:

    • multixact_members_cache_size

    • multixact_offsets_cache_size

  • Fixed an issue to detect and abort stuck major version upgrades.

  • Fixed an issue in hash join that could lead to increased memory consumption.

  • Improved the logical replication performance.

  • Upgraded the GDAL version to 3.4.3.

  • Upgraded the GEOS version to 3.10.3.

  • Updated the PLV8 version to 3.0.0.

  • Updated the PostGIS extension to version 3.2.3.

  • Fixed an issue with st_orientedenvelope that caused it to loop with a 1-D input to return 0.

  • Fixed an issue where the connection to SQL Server using tds_fdw fails.

PostgreSQL 14.4

This release of Aurora PostgreSQL is compatible with PostgreSQL 14.4. For more information about the improvements in PostgreSQL 14.4, see PostgreSQL release 14.4.

Aurora PostgreSQL 14.4.4, November 17, 2022

High priority stability enhancements

  • Fixed an issue that can cause increased network traffic when a writer instance transmits logs to a replica instance.

Aurora PostgreSQL 14.4.0, October 13, 2022

High priority stability enhancements

  • Fixed an issue where PLV8 crashed during a JavaScript error scenario.

  • Fixed an issue where PLV8 crashed when trying to acquire a semaphore to execute the next task.

  • Fixed an issue where scaling an Aurora Serverless v2 instance gets stuck if VACUUM is running.

General stability enhancements

  • This release contains all of the fixes, features, and improvements present in PostgreSQL release 14.3.

  • Fixed a bug where Aurora PostgreSQL can't file the relfilenode.

  • Fixed a database restart issue when a plan gets invaliated but the engine still checks if it is valid.

  • Fixed a stuck scaling issue when the current scaling event times out.

  • Upgraded the PLV8 extension to version 3.0.0.

  • Upgraded the PostGIS extension to version 3.2.3.

  • Fixed an issue where extended query messages might be lost during zero-downtime patching (ZDP) causing the extended query to hang after the ZDP completion.

PostgreSQL 14.3

This release of Aurora PostgreSQL is compatible with PostgreSQL 14.3. For more information about the improvements in PostgreSQL 14.3, see PostgreSQL release 14.3.

Aurora PostgreSQL 14.3.4, November 17, 2022

High priority stability enhancements

  • Fixed an issue that can cause increased network traffic when a writer instance transmits logs to a replica instance.

Aurora PostgreSQL 14.3.3, October 13, 2022

High priority stability enhancements

  • Fixed a PLV8 issue where the base parameter doesn't get loaded properly into the memory.

  • Fixed an issue where scaling an Aurora Serverless v2 instance gets stuck if VACUUM is running.

General stability enhancements

  • Fixed a bug where Aurora PostgreSQL can't file the relfilenode.

  • Fixed a database restart issue when a plan gets invaliated but the engine still checks if it is valid.

  • Fixed a stuck scaling issue when the current scaling event times out.

  • Upgraded the PostGIS extension to version 3.1.7.

  • Fixed an issue where extended query messages might be lost during zero-downtime patching (ZDP) causing the extended query to hang after the ZDP completion.

Aurora PostgreSQL 14.3.1, July 6, 2022

Critical stability enhancements

  • Fixed an issue that could cause periods of unavailability during a storage node restart.

High priority stability enhancements

  • Fixed an error handling issue related to out-of-memory conditions which could result in brief periods of unavailability.

  • Fixed an issue when the connection to SQL Server fails using the TDS_FDW extension to query a foreign table.

  • Fixed an issue that caused connections using the provided root certificate to fail.

  • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

Aurora PostgreSQL 14.3.0, June 21, 2022

New features

Additional improvements and enhancements

  • Contains all of the fixes, features, and improvements present in Aurora PostgreSQL 13.7.

  • Backported the following bug fix from the PostgreSQL 14.4 release: Reverted changes to CONCURRENTLY that "sped up" Xmin advance to prevent Index Corruption with the CREATE INDEX CONCURRENTLY / REINDEX CONCURRENTLY commands.

  • This release supports lo extension version 1.1.

  • This release supports old_snapshot extension version 1.0.

  • This release supports EBCDIC collations for the mainframe modernization efforts. For more information, see Aurora PostgreSQL collations for EBCDIC and other mainframe migrations in the Amazon Aurora User Guide.

  • Updated the following extensions:

    • amcheck to version 1.3

    • btree_gist to version 1.6

    • cube to version 1.5

    • hll to version 2.16

    • hstore to version 1.8

    • intarray to version 1.5

    • log_fdw to version 1.3

    • oracle_fdw to version 2.4.0

    • pg_hint_plan to version 1.4

    • pg_partman to version 4.6.0

    • pg_repack to version 1.4.7

    • pg_stat_statements to version 1.9

    • pg_trgm to version 1.6

    • pgaudit to version 1.6.1

    • pgrouting to version 3.2.0

    • pgtap to version 1.2.0

    • postgres_fdw to version 1.1

PostgreSQL 13.8

This release of Aurora PostgreSQL is compatible with PostgreSQL 13.8. For more information about the improvements in PostgreSQL 13.8, see PostgreSQL release 13.8.

Aurora PostgreSQL 13.8.0, November 09, 2022

High priority stability enhancements

  • Fixed an Aurora Serverless v2 scaling issue.

  • Fixed an issue which caused Aurora Serverless v2 shrink failure.

General stability enhancements

  • This release contains all of the fixes, features, and improvements present in PostgreSQL release 13.7.

  • Improved buffer cache scavenging when the buffer cache is in duress.

  • Fixed an issue in Database Activity Streams that leads to high memory consumption.

  • Fixed an issue that caused DB instance restarts.

  • Fixed an issue where a DB instance restarts recursively while generating monitoring metrics during a crash.

  • Fixed an issue where a DB instance restarted during performance metric collection.

  • Fixed an issue where an attempt to connect to the database would fail with SSLV3_ALERT_CERTIFICATE_UNKNOWN.

  • Improved the error reporting in case of an inconsistent B-tree index.

  • Improved the diagnostic logging around setting invalid hint bits.

  • Fixed an issue where autovacuum would incorrectly skip tables.

  • Improved the logical replication prefetching.

  • Fixed a durability issue in the GIN indexes.

  • Provided options to configure MultiXact SLRU cache. By setting the following parameters (GUCs), customers will now be able to change the MultiXact SLRU cache sizes:

    • multixact_members_cache_size

    • multixact_offsets_cache_size

  • Fixed an issue to detect and abort stuck major version upgrades.

  • Fixed an issue in hash join that could lead to increased memory consumption.

  • Improved the logical replication performance.

  • Upgraded the GDAL version to 3.4.3.

  • Upgraded the GEOS version to 3.10.3.

  • Updated the PLV8 version to 3.0.0.

  • Updated the PostGIS extension to version 3.2.3.

  • Fixed an issue with st_orientedenvelope that caused it to loop with a 1-D input to return 0.

  • Fixed an issue where the connection to SQL Server using tds_fdw fails.

PostgreSQL 13.7

This release of Aurora PostgreSQL is compatible with PostgreSQL 13.7. For more information about the improvements in PostgreSQL 13.7, see PostgreSQL 13.7.

Aurora PostgreSQL 13.7.4, November 17, 2022

High priority stability enhancements

  • Fixed an issue that can cause increased network traffic when a writer instance transmits logs to a replica instance.

Aurora PostgreSQL 13.7.3, October 13, 2022

High priority stability enhancements

  • Fixed a PLV8 issue where the base parameter doesn't get loaded properly into the memory.

  • Fixed an issue where scaling an Aurora Serverless v2 instance gets stuck if VACUUM is running.

General stability enhancements

  • Fixed a bug where Aurora PostgreSQL can't file the relfilenode.

  • Fixed a database restart issue when a plan gets invaliated but the engine still checks if it is valid.

  • Fixed a stuck scaling issue when the current scaling events time out.

  • Upgraded the PostGIS extension to version 3.1.7.

  • Fixed an issue where extended query messages might be lost during zero-downtime patching (ZDP) causing the extended query to hang after the ZDP completion.

Aurora PostgreSQL 13.7.1, July 6, 2022

Critical stability enhancements

  • Fixed an issue that could cause periods of unavailability during a storage node restart.

High priority stability enhancements

  • Fixed an error handling issue related to out-of-memory conditions which could result in brief periods of unavailability.

  • Fixed an issue when the connection to SQL Server fails using the TDS_FDW extension to query a foreign table.

  • Fixed an issue that caused connections using the provided root certificate to fail.

  • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

Aurora PostgreSQL 13.7.0, June 9, 2022

New features

Critical updates

  • Fixed a replay crash due to an LSN mismatch.

  • Fixed the S3 extension to prevent invalid region injection.

High stability updates

  • Fixed multiple issues related to out of memory conditions which could result in brief periods of unavailability.

  • Fixed an Aurora Serverless v2 scaling issue.

General enhancements

  • Fixed a lock contention crash during an Aurora Serverless v1 scaling event.

  • Fixed an issue where logical replication becomes stuck after a restart.

  • Fixed multiple issues that could lead to brief periods of unavailability.

  • Fixed a crash in pg_cron due to a task still running but being unscheduled.

  • Fixed, during redo, an invalid page hit on the Generic Redo for GENERIC_XLOG_FULL_PAGE_DATA. This happens due to a timing hole between generating the log record and then writing the metadata for the record on the RW node and the RO node replays between those operations.

  • Improved the query performance by supporting parallel workers.

  • Upgraded the plugin wal2json version to 2.4.

  • Upgraded the pglogical extension to version 2.4.1.

PostgreSQL 13.6

This release of Aurora PostgreSQL is compatible with PostgreSQL 13.6. For more information about the improvements in PostgreSQL 13.6, see PostgreSQL release 13.6.

Aurora PostgreSQL 13.6.5, October 18, 2022

High priority enhancements

  • Fixed an issue where Amazon Aurora Serverless v2 scaling may get blocked if VACUUM is running.

  • Fixed an issue where Amazon Aurora Serverless v2 scaling may get blocked on Aurora replicas.

General enhancements

  • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

  • Updated the PostGIS extension to version 3.1.7.

Aurora PostgreSQL 13.6.4, July 18, 2022

Security enhancements

  • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

Critical enhancements

  • Fixed an issue during a storage node restart that could result in periods of unavailability.

High priority stability enhancements

  • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

  • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

  • Fixed a defect where the validation of cached plans may lead to a database restart when the plan was previously invalidated.

Aurora PostgreSQL 13.6.3, June 2, 2022

Security enhancements

  • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

High priority stability updates

  • Fixed an issue that can lead to a blocked scale operation when a COPY command is in progress in Amazon Aurora Serverless v2.

  • Fixed an issue that can cause a restart of the database when dropping or truncating tables in Amazon Aurora Serverless v2.

  • Fixed an issue in the apg_prewarm extension that can lead to a blocked scale operation in Amazon Aurora Serverless v2.

  • Fixed an issue in the dynamic shared memory allocation that can lead to a blocked scale operation in Amazon Aurora Serverless v2.

  • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v2.

  • Fixed an issue where a minor version ugprade may get blocked when there is a SQL view that refers to the checksum() function in Babelfish for Aurora PostgreSQL.

  • Fixed an issue in apg_plan_mgmt that can cause a restart when Query Plan Management (QPM) is enabled.

Aurora PostgreSQL 13.6.2, May 12, 2022

High priority stability enhancements

  • Fixed an issue that causes upgrades to fail when Babelfish for Aurora PostgreSQL is enabled.

  • Fixed an issue that causes scaling in Aurora Serverless v2 to fail.

General enhancements

  • Fixed an issue that could cause unavailability due to improper locking of shared memory.

Aurora PostgreSQL 13.6.1, April 27, 2022

High priority stability enhancements

  • Fixed an issue that could cause incorrect WriteIOPS reporting in the Amazon console.

  • Fixed an issue that could cause unavailability after removal of a read node from a cluster.

General enhancements

  • Fixed an issue that could cause an engine restart during periods of low free memory.

Aurora PostgreSQL 13.6.0, March 29, 2022

New features

  • Added support for the tds_fdw extension version 2.0.2.

High priority stability enhancements

  • Fixed multiple issues that may result in unavailability of a read node.

  • Fixed an issue that may result in a read node being unable to replay WAL requiring the replication slot to be dropped and resynchronized.

  • Fixed an issue that could cause excess storage use due to files not being properly closed.

General enhancements

  • Fixed a small memory leak on read nodes when commit_ts is set.

  • Fixed an issue that caused Performance Insights to show "Unknown wait event".

  • Fixed an issue that could cause an import from S3 to fail when using the S3 Extension.

  • Fixed multiple issues that could result in periods of unavailability when using apg_plan_mgmt.

  • Fixed multiple issues that could result in periods of unavailability when QPM is enabled.

PostgreSQL 13.5

This release of Aurora PostgreSQL is compatible with PostgreSQL 13.5. For more information about the improvements in PostgreSQL 13.5, see PostgreSQL release 13.5.

Aurora PostgreSQL 13.5.5, October 18, 2022

General enhancements

  • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

  • Updated the PostGIS extension to version 3.1.7.

Aurora PostgreSQL 13.5.4, July 20, 2022

Security enhancements

  • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

Critical enhancements

  • Fixed an issue during a storage node restart that could result in periods of unavailability.

High stability enhancements

  • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

  • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

  • Fixed an issue that could cause excess storage use due to files not being properly closed.

  • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v2.

  • Fixed an issue that caused Performance Insights to show "Unknown wait event".

Aurora PostgreSQL 13.5.3, April 13, 2022

Security enhancements

  • Additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

General enhancements

  • Fixed a bug that could cause an engine restart during periods of low free memory.

Aurora PostgreSQL 13.5.1, March 3, 2022

Security enhancements

  • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

Aurora PostgreSQL 13.5.0, February 25, 2022

High priority stability enhancements

  • Fixed a bug where logical replication may hang resulting in the replay falling behind on the read node. The instance may eventually restart.

Additional improvements and enhancements

  • Added the Buffers: shared hit metric to the Explain output.

  • Fixed a buffer cache bug that could cause brief periods of unavailability.

  • Fixed a bug in the apg_plan_mgmt extension where an index based plan was not being enforced.

  • Fixed a bug in the pg_logical extension that could cause brief periods of unavailability due to improper handling of NULL arguments.

  • Fixed a bug that could cause brief periods of unavailability due to reading uninitialized pages.

  • Fixed an issue where orphaned files caused major version upgrades to fail.

  • Fixed incorrect Aurora Storage Daemon log write metrics.

  • Fixed multiple bugs that could result in WAL replay falling behind and eventually causing the reader instances to restart.

  • Improved the Aurora buffer cache page validation on reads.

  • Improved the Aurora storage metadata validation.

This version also includes the following change:

  • The pg_cron extension is updated to 1.4.1

For information about extensions and modules, see Extensions supported for Aurora PostgreSQL 13.

PostgreSQL 13.4

This release of Aurora PostgreSQL is compatible with PostgreSQL 13.4. For more information about the improvements in PostgreSQL 13.4, see PostgreSQL release 13.4.

Aurora PostgreSQL 13.4.5, October 18, 2022

General enhancements

  • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

  • Updated the PostGIS extension to version 3.1.7.

Aurora PostgreSQL 13.4.4, July 6, 2022

Security enhancements

  • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

General enhancements

  • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

  • Fixed an issue that could cause excess storage use due to files not being properly closed.

  • Fixed an issue that could cause a restart of the postmaster process in Amazon Aurora Serverless v2.

  • Fixed an issue that could cause Performance Insights to display "Unknown wait event".

  • Fixed an issue that could result in periods of unavailability due to the existence of duplicate relation files.

Aurora PostgreSQL 13.4.2, April 12, 2022

Security enhancements

  • Additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

General enhancements

  • Fixed a buffer cache bug that could cause brief periods of unavailability.

Aurora PostgreSQL 13.4.1

Security enhancements

  • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

Aurora PostgreSQL 13.4.0

New features

  • This version supports Babelfish 1.0.0 which extends your Amazon Aurora PostgreSQL database with the ability to accept database connections from Microsoft SQL Server clients. For more information, see Working with Babelfish for Aurora PostgreSQL.

Critical stability enhancements

  • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

High priority stability enhancements

  • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

  • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

Additional improvements and enhancements

  • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

  • Fixed an issue where read queries may time out on read nodes during the replay of lazy truncation triggered by vacuum on the write node.

  • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

  • Fixed an issue where the aurora_postgres_replica_status() function returned stale or lagging CPU stats.

  • Fixed an issue where the role rds_superuser did not have permission to execute the pg_stat_statements_reset() function.

  • Fixed an issue with the apg_plan_mgmt extension where the planning and execution times were reported as 0.

  • Removed support for DES, 3DES and RC4 cipher suites.

  • Updated the PostGIS extension to version 3.1.4.

  • Updated the pgrouting extension to 3.1.3.

  • Updated the pglogical extension to 2.4.0.

  • Added support for the following SPI module extensions:

    • autoinc version 1.0

    • insert_username version 1.0

    • moddatetime version 1.0

    • refint version 1.0

  • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

  • Fixed an out of memory crash issue with Aurora storage daemon that leads to writer node restart. This also reduces the overall system memory consumption.

PostgreSQL 13.3

This release of Aurora PostgreSQL is compatible with PostgreSQL 13.3. For more information about the improvements in PostgreSQL 13.3, see PostgreSQL release 13.3.

Aurora PostgreSQL 13.3.4, July 14, 2022

Security enhancements

  • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

High priority stability enhancements

  • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

  • Fixed an issue that could cause excess storage use due to files not being properly closed.

  • Fixed an issue that caused Performance Insights to show "Unknown wait event".

Aurora PostgreSQL 13.3.3, April 7, 2022

Security enhancements

  • Includes additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

Aurora PostgreSQL 13.3.2

Security enhancements

  • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

  • Backpatched postgis to PostGIS 3.0.3. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

Aurora PostgreSQL 13.3.1

Releases and patches

    Security enhancements

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Critical stability enhancements

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability enhancements

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue with the apg_plan_mgmt extension where the planning and execution times were reported as 0.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue with the apg_plan_mgmt extension where plan outline on a partitioned table did not enforce an index-based plan.

    • Fixed an issue where orphaned files caused failed translations in read codepaths during or after a major version upgrade.

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    • Fixed an out of memory crash issue with Aurora storage daemon that leads to writer node restart. This also reduces the overall system memory consumption.

    Aurora PostgreSQL 13.3.0

    New features

    Critical stability enhancements

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    Additional improvements and enhancements

    • Contains all of the fixes, features, and improvements present in PostgreSQL 12.7, Aurora PostgreSQL 4.2

    • Contains several improvements that were announced for PostgreSQL releases 13.0, 13.1, 13.2 and 13.3

    • Instance type R4 was deprecated.

    • Updated the following extensions:

      • hll to version 2.15.

      • hstore to version 1.7.

      • intarray to version 1.3.

      • log_fdw to version 1.2.

      • ltree to version 1.2.

      • pg_hint_plan to version 1.3.7.

      • pg_repack to version 1.4.6.

      • pg_stat_statements to version 1.8.

      • pg_trgm to version 1.5.

      • pgaudit to version 1.5.

      • pglogical to version 2.3.3.

      • pgrouting to version 3.1.0

      • plcoffee to version 2.3.15.

      • plls to version 2.3.15.

      • plv8 to version 2.3.15.

    PostgreSQL 12.12

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.12. For more information about the improvements in PostgreSQL 12.12, see PostgreSQL release 12.12.

    Aurora PostgreSQL 12.12.0, November 09, 2022

    High priority stability enhancements

    • Fixed an Aurora Serverless v2 scaling issue.

    General stability enhancements

    • This release contains all of the fixes, features, and improvements present in PostgreSQL release 12.11.

    • Added support for rds_superusers to execute CREATE OPERATOR CLASS, CREATE OPERATOR FAMILY and ALTER OPERATOR FAMILY, which are available in the higher versions.

    • Improved buffer cache scavenging when the buffer cache is in duress.

    • Fixed an issue in Database Activity Streams that leads to high memory consumption.

    • Fixed an issue that caused DB instance restarts.

    • Fixed an issue where a DB instance restarts recursively while generating monitoring metrics during a crash.

    • Fixed an issue where a DB instance restarted during performance metric collection.

    • Fixed an issue where an attempt to connect to the database would fail with SSLV3_ALERT_CERTIFICATE_UNKNOWN.

    • Improved the diagnostic logging around setting invalid hint bits.

    • Fixed an issue where autovacuum would incorrectly skip tables.

    • Improved the logical replication prefetching.

    • Fixed a durability issue in the GIN indexes.

    • Fixed an issue to detect and abort stuck major version upgrades.

    • Fixed an issue in hash join that could lead to increased memory consumption.

    • Improved the logical replication performance.

    • Upgraded the GDAL version to 3.4.3.

    • Upgraded the GEOS version to 3.10.3.

    • Updated the PostGIS extension to version 3.2.3.

    • Fixed an issue with st_orientedenvelope that caused it to loop with a 1-D input to return 0.

    • Fixed an issue where the connection to SQL Server using tds_fdw fails.

    PostgreSQL 12.11

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.11. For more information about the improvements in PostgreSQL 12.11, see PostgreSQL release 12.11.

    Aurora PostgreSQL 12.11.4, November 17, 2022

    High priority stability enhancements

    • Fixed an issue that can cause increased network traffic when a writer instance transmits logs to a replica instance.

    Aurora PostgreSQL 12.11.3, October 13, 2022

    High priority stability enhancements

    • Fixed a PLV8 issue where the base parameter doesn't get loaded properly into the memory.

    • Fixed an issue where scaling an Aurora Serverless v2 instance gets stuck if VACUUM is running.

    General stability enhancements

    • Fixed a bug where Aurora PostgreSQL can't file the relfilenode.

    • Fixed a stuck scaling issue when the current scaling event times out.

    • Upgraded the PostGIS extension to version 3.1.7.

    • Fixed an issue where extended query messages might be lost during zero-downtime patching (ZDP) causing the extended query to hang after the ZDP completion.

    Aurora PostgreSQL 12.11.1, July 6, 2022

    Critical stability enhancements

    • Fixed an issue that could cause periods of unavailability during a storage node restart.

    High priority stability enhancements

    • Fixed an error handling issue related to out-of-memory conditions which could result in brief periods of unavailability.

    • Fixed an issue when the connection to SQL Server fails using the TDS_FDW extension to query a foreign table.

    • Fixed an issue that caused connections using the provided root certificate to fail.

    • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

    Aurora PostgreSQL 12.11.0, June 9, 2022

    New features

    Critical updates

    • Fixed a replay crash due to an LSN mismatch.

    • Fixed the S3 extension to prevent invalid region injection.

    High stability updates

    • Fixed multiple issues related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an Aurora Serverless v2 scaling issue.

    General stability updates

    • Fixed a lock contention crash during an Aurora Serverless v1 scaling event.

    • Fixed an issue where logical replication becomes stuck after a restart.

    • Fixed multiple issues that could lead to brief periods of unavailability.

    • Fixed a crash in pg_cron due to a task still running but being unscheduled.

    • Fixed, during redo, an invalid page hit on the Generic Redo for GENERIC_XLOG_FULL_PAGE_DATA. This happens due to a timing hole between generating the log record and then writing the metadata for the record on the RW node and the RO node replays between those operations.

    • Improved the query performance by supporting parallel workers.

    • Upgraded the plugin wal2json version to 2.4.

    • Upgraded the pglogical extension to version 2.4.1.

    PostgreSQL 12.10

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.10. For more information about the improvements in PostgreSQL 12.10, see PostgreSQL release 12.10.

    Aurora PostgreSQL 12.10.4, July 18, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Critical enhancements

    • Fixed an issue during a storage node restart that could result in periods of unavailability.

    High stability enhancements

    • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

    • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

    • Fixed a defect where the validation of cached plans may lead to a database restart when the plan was previously invalidated.

    Aurora PostgreSQL 12.10.1, April 27, 2022

    High priority stability enhancements

    • Fixed an issue that could cause incorrect WriteIOPS reporting in the Amazon console.

    • Fixed an issue that could cause unavailability after removal of a read node from a cluster.

    General enhancements

    • Fixed an issue that could cause an engine restart during periods of low free memory.

    Aurora PostgreSQL 12.10.0, March 29, 2022

    High priority stability enhancements

    • Fixed multiple issues that may result in unavailability of a read node.

    • Fixed an issue that may result in a read node being unable to replay WAL requiring the replication slot to be dropped and resynchronized.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    General enhancements

    • Fixed a small memory leak on read nodes when commit_ts is set.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    • Fixed an issue that could cause an import from S3 to fail when using the S3 Extension.

    • Fixed multiple issues that could result in periods of unavailability when using apg_plan_mgmt.

    • Fixed multiple issues that could result in periods of unavailability when QPM is enabled.

    PostgreSQL 12.9

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.9. For more information about the improvements in PostgreSQL 12.9, see PostgreSQL release 12.9.

    Aurora PostgreSQL 12.9.4, July 20, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Critical enhancements

    • Fixed an issue during a storage node restart that could result in periods of unavailability.

    High stability enhancements

    • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

    • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v2.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    Aurora PostgreSQL 12.9.3, April 13, 2022

    Security enhancements

    • Additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    General enhancements

    • Fixed a bug that could cause an engine restart during periods of low free memory.

    Aurora PostgreSQL 12.9.1

    Security enhancements

    • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 12.9.0

    Critical stability enhancements

    • Fixed a bug where logical replication may hang resulting in the replay falling behind on the read node. The instance may eventually restart.

    Additional improvements and enhancements

    • Fixed a buffer cache bug that could cause brief periods of unavailability.

    • Fixed a bug in the apg_plan_mgmt extension where an index based plan was not being enforced.

    • Fixed a bug in the pg_logical extension that could cause brief periods of unavailability due to improper handling of NULL arguments.

    • Fixed an issue where orphaned files caused major version upgrades to fail.

    • Fixed incorrect Aurora Storage Daemon log write metrics.

    • Fixed multiple bugs that could result in WAL replay falling behind and eventually causing the reader instances to restart.

    • Improved the Aurora buffer cache page validation on reads.

    • Improved the Aurora storage metadata validation.

    • Updated the pg_cron extension to v1.4.

    • Updated the pg_hint_pan extension to v1.3.7.

    PostgreSQL 12.8

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.8. For more information about the improvements in PostgreSQL 12.8, see PostgreSQL release 12.8.

    Aurora PostgreSQL 12.8.4, July 6, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    General enhancements

    • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that could cause a restart of the postmaster process in Amazon Aurora Serverless v2.

    • Fixed an issue that could cause Performance Insights to display "Unknown wait event".

    • Fixed an issue that could result in periods of unavailability due to the existence of duplicate relation files.

    Aurora PostgreSQL 12.8.2, April 12, 2022

    Security enhancements

    • Additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    General enhancements

    • Fixed a buffer cache bug that could cause brief periods of unavailability.

    Aurora PostgreSQL 12.8.1

    Security enhancements

    • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 12.8.0

    Critical stability enhancements

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability enhancements

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue where read queries may time out on read nodes during the replay of lazy truncation triggered by vacuum on the write node.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue where the aurora_postgres_replica_status() function returned stale or lagging CPU stats.

    • Fixed an issue where the role rds_superuser did not have permission to execute the pg_stat_statements_reset() function.

    • Fixed an issue with the apg_plan_mgmt extension where the planning and execution times were reported as 0.

    • Removed support for DES, 3DES and RC4 cipher suites.

    • Updated PostGIS extension to version 3.1.4.

    PostgreSQL 12.7, Aurora PostgreSQL 4.2

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.7. For more information about the improvements in PostgreSQL 12.7, see PostgreSQL release 12.7.

    Aurora PostgreSQL 12.7.4, July 14, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    High priority stability enhancements

    • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    Aurora PostgreSQL 4.2.3, April 7, 2022

    Security enhancements

    • Additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 4.2.2

    Security enhancements

    • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched the PostgreSQL community fix for CVE-2021-3677: “Memory disclosure in certain queries”. For more information, see CVE-2021-3677

    • Backpatched postgis to PostGIS 3.0.3. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 4.2.1

    Critical stability enhancements

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability enhancements

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue with the apg_plan_mgmt extension where planning and execution time were reported as 0.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue with the apg_plan_mgmt extension where plan outline on a partitioned table did not enforce an index-based plan.

    • Fixed an issue where orphaned files caused failed translations in read codepaths during or after major version upgrade.

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    • Fixed an out of memory crash issue with Aurora storage daemon that leads to writer node restart. This also reduces the overall system memory consumption.

    Aurora PostgreSQL 4.2.0

    New features

    • Added support for the oracle_fdw extension version 2.3.0.

    High priority stability enhancements

    • Fixed an issue where creating a database from an existing template database with tablespace resulted in an error with the message ERROR: could not open file pg_tblspc/...: No such file or directory.

    • Fixed an issue where, in rare cases, an Aurora replica may be unable to start when a large number of PostgreSQL subtransactions (i.e. SQL savepoints) have been used.

    • Fixed an issue where, in rare circumstances, read results may be inconsistent for repeated read requests on replica nodes.

    Additional improvements and enhancements

    • Upgraded OpenSSL to 1.1.1k.

    • Reduced CPU usage and memory consumption of the WAL apply process on Aurora replicas for some workloads.

    • Improved safety checks in the write path to detect incorrect writes to metadata.

    • Improved security by removing 3DES and other older ciphers for SSL/TLS connections.

    • Fixed an issue where a duplicate file entry can prevent the Aurora PostgreSQL engine from starting up.

    • Fixed an issue that could cause temporary unavailability under heavy workloads.

    • Added back ability to use a leading forward slash in the S3 path during S3 import.

    • Added Graviton support for oracle_fdw extension version 2.3.0.

    • Changed the following extensions:

      • Updated the Orafce extension to version 3.16.

      • Updated the pg_partman extension to version 4.5.1.

      • Updated the pg_cron extension to version 1.3.1.

      • Updated the postgis extension to version 3.0.3.

    PostgreSQL 12.6, Aurora PostgreSQL 4.1 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.6. For more information about the improvements in PostgreSQL 12.6, see PostgreSQL release 12.6.

    Aurora PostgreSQL 4.1.2, April 7, 2022

    Security enhancements

    • Additional modifications to the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 4.1.1

    Security enhancements

    • Modified the pg_cron extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched the PostgreSQL community fix for CVE-2021-3677: “Memory disclosure in certain queries”. For more information, see CVE-2021-3677

    • Backpatched pg_partman to 4.4.0. This is a pg_partman fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched postgis to PostGIS 3.0.2. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched an input validation error in the log_fdw extension function parameters.

    Aurora PostgreSQL 4.1.0

    New features

    High priority stability enhancements

    • Fixed a bug in the pglogical extension that could lead to data inconsistency for inbound replication.

    • Fixed a bug where in rare cases a reader had inconsistent results when it restarted while a transaction with more than 64 subtransactions was being processed.

    • Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    • Fixed a bug where the database could not be started when there were many relations in memory-constrained environments.

    • Fixed a bug in the apg_plan_mgmt extension that could cause brief periods of unavailability due to an internal buffer overflow.

    • Fixed a bug on reader nodes that could cause brief periods of unavailability during WAL replay.

    • Fixed a bug in the rds_activity_stream extension that caused an error during startup when attempting to log audit events.

    • Fixed bugs in the aurora_replica_status function where rows were sometimes partially populated and some values such as Replay Latency, and CPU usage were always 0.

    • Fixed a bug where the database engine would attempt to create shared memory segments larger than the instance total memory and fail repeatedly. For example, attempts to create 128 GiB shared buffers on a db.r5.large instance would fail. With this change, requests for total shared memory allocations larger than the instance memory allow setting the instance to incompatible parameters.

    • Added logic to clean up unnecessary pg_wal temporary files on a database startup.

    • Fixed a bug that could lead to outbound replication synchronization errors after a major version upgrade.

    • Fixed a bug that reported ERROR: rds_activity_stream stack item 2 not found on top - cannot pop when attempting to create the rds_activity_stream extension.

    • Fixed a bug that could cause the error failed to build any 3-way joins in a correlated IN subquery under an EXISTS subquery.

    • Backported the following performance improvement from the PostgreSQL community: pg_stat_statements: add missing check for pgss_enabled().

    • Fixed a bug that could cause upgrades to Aurora PostgreSQL 12.x to fail due to the inability to open the pg_control file.

    • Fixed a bug that could cause brief periods of unavailability due to running out of memory when creating the postgis extension with pgAudit enabled.

    • Backported the following bug fix from the PostgreSQL community: Fix use-after-free bug with AfterTriggersTableData.storeslot.

    • Fixed a bug when using outbound logical replication to synchronize changes to another database that could fail with an error message like ERROR: could not map filenode "base/16395/228486645" to relation OID.

    • Fixed a bug that could cause a brief period of unavailability when aborting a transaction.

    • Fixed a bug that caused no ICU collations to be shown in the pg_collation catalog table after creating a new Aurora PostgreSQL 12.x instance. This issue does not affect upgrading from an older version.

    • Fixed a bug where the rds_ad role wasn't created after upgrading from a version of Aurora PostgreSQL that doesn't support Microsoft Active Directory authentication.

    • Added btree page checks to detect tuple metadata inconsistency.

    • Fixed a bug in asynchronous buffer reads that could cause brief periods of unavailability on reader nodes during WAL replay.

    • Fixed a bug where reading a TOAST value from disk could cause a brief period of unavailability.

    • Fixed a bug that caused brief periods of unavailability when attempting to fetch a tuple from and index scan.

    PostgreSQL 12.4, Aurora PostgreSQL 4.0 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 12.4. For more information about the improvements in PostgreSQL 12.4, see PostgreSQL release 12.4.

    Aurora PostgreSQL 4.0.5

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched the PostgreSQL community fix for CVE-2021-3677: “Memory disclosure in certain queries”. For more information, see CVE-2021-3677

    • Backpatched postgis to PostGIS 3.0.2. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched an input validation error in the log_fdw extension function parameters.

    Aurora PostgreSQL 4.0.2

    High priority stability enhancements

    • Fixed a bug where a reader node might render an extra or missing row if the reader restarted while the writer node is processing a long transaction with more than 64 subtransactions.

    • Fixed a bug that can cause vacuum to block on GiST indexes.

    • Fixed a bug where after upgrade to PostgreSQL 12, vacuum can fail on the system table pg_catalog.pg_shdescription with the following error. ERROR: found xmin 484 from before relfrozenxid

    Additional improvements and enhancements

    • Fixed a bug that could lead to intermittent unavailability due to a race condition when handling responses from storage nodes.

    • Fixed a bug that could lead to intermittent unavailability due to the rotation of network encryption keys.

    • Fixed a bug that could lead to intermittent unavailability due to heat management of the underlying storage segments.

    • Fixed a bug where a large S3 import with thousands of clients can cause one or more of the import clients to stop responding.

    • Removed a restriction that prevented setting configuration variable strings that contained brazil.

    • Fixed a bug that could lead to intermittent unavailability if a reader node runs a query that access many tables while the writer node is acquiring exclusive locks on all of the same tables.

    Aurora PostgreSQL 4.0.1

    New features

    Critical stability enhancements

    • Fixed a bug that caused a read replica to unsuccessfully restart repeatedly in rare cases.

    • Fixed a bug where a cluster became unavailable when attempting to create more than 16 read replicas or Aurora global database secondary Amazon Web Services Regions. The cluster became available again when the new read replica or secondary Amazon Web Services Region was removed.

    Additional improvements and enhancements

    • Fixed a bug that when under heavy load, snapshot import, COPY import, or Amazon S3 import stopped responding in rare cases.

    • Fixed a bug where a read replica might not join the cluster when the writer was very busy with a write-intensive workload.

    • Fixed a bug where a cluster could be unavailable briefly when a high-volume S3 import was running.

    • Fixed a bug that caused a cluster to take several minutes to restart if a logical replication stream was terminated while handling many complex transactions.

    • Fixed the Just-in-Time (JIT) compilation, which was incorrectly enabled by default in Aurora PostgreSQL 4.0.0.

    • Disallowed the use of both Amazon Identity and Access Management (IAM) and Kerberos authentication for the same user.

    Aurora PostgreSQL 4.0.0

    New features

    Additional improvements and enhancements

    • Contains several improvements that were announced for PostgreSQL releases 12.0, 12.1, 12.2, 12.3, and 12.4.

    • Contains all fixes, features, and improvements present in PostgreSQL 11.9, Aurora PostgreSQL 3.4 .

    • Backported fixes for the following PostgreSQL community security issues:

    • Updated the following extensions:

      • address_standardizer to version 3.0.2

      • address_standardizer_data_us to version 3.0.2

      • amcheck to version 1.2

      • citext to version 1.6

      • hll to version 2.14

      • hstore to version 1.6

      • ip4r to version 2.4

      • pg_repack to version 1.4.5

      • pg_stat_statements to version 1.7

      • pgaudit to version 1.4

      • pglogical to version 2.3.2

      • pgrouting to version 3.0.3

      • plv8 to version 2.3.14

      • postGIS to version 3.0.2

      • postgis_tiger_geocoder to version 3.0.2

      • postgis_topology to version 3.0.2

    PostgreSQL 11.17

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.17. For more information about the improvements in PostgreSQL 11.17, see PostgreSQL release 11.17.

    Aurora PostgreSQL 11.17.0, November 09, 2022

    High priority stability enhancements

    • Fixed an Aurora Serverless v2 scaling issue.

    General stability enhancements

    • This release contains all of the fixes, features, and improvements present in PostgreSQL release 11.16.

    • Improved buffer cache scavenging when the buffer cache is in duress.

    • Fixed an issue in Database Activity Streams that leads to high memory consumption.

    • Fixed an issue that caused DB instance restarts.

    • Fixed an issue where a DB instance restarts recursively while generating monitoring metrics during a crash.

    • Fixed an issue where a DB instance restarted during performance metric collection.

    • Fixed an issue where an attempt to connect to the database would fail with SSLV3_ALERT_CERTIFICATE_UNKNOWN.

    • Improved the diagnostic logging around setting invalid hint bits.

    • Fixed an issue where autovacuum would incorrectly skip tables.

    • Improved the logical replication prefetching.

    • Fixed a durability issue in the GIN indexes.

    • Fixed an issue to detect and abort stuck major version upgrades.

    • Fixed an issue in hash join that could lead to increased memory consumption.

    • Improved the logical replication performance.

    • Upgraded the GDAL version to 3.4.3.

    • Upgraded the GEOS version to 3.10.3.

    • Updated the PostGIS extension to version 3.2.3.

    • Fixed an issue with st_orientedenvelope that caused it to loop with a 1-D input to return 0.

    • Fixed an issue where the connection to SQL Server using tds_fdw fails.

    PostgreSQL 11.16

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.16. For more information about the improvements in PostgreSQL 11.16, see PostgreSQL release 11.16.

    Aurora PostgreSQL 11.16.4, November 17, 2022

    High priority stability enhancements

    • Fixed an issue that can cause increased network traffic when a writer instance transmits logs to a replica instance.

    Aurora PostgreSQL 11.16.3, October 13, 2022

    High priority stability enhancements

    • Fixed a PLV8 issue where the base parameter doesn't get loaded properly into the memory.

    • Fixed an issue where scaling an Aurora Serverless v2 instance gets stuck if VACUUM is running.

    General stability enhancements

    • Fixed a bug where Aurora PostgreSQL can't file the relfilenode.

    • Fixed a stuck scaling issue when the current scaling event times out.

    • Upgraded the PostGIS extension to version 3.1.7.

    • Fixed an issue where extended query messages might be lost during zero-downtime patching (ZDP) causing the extended query to hang after the ZDP completion.

    Aurora PostgreSQL 11.16.1, July 6, 2022

    Critical stability enhancements

    • Fixed an issue that could cause periods of unavailability during a storage node restart.

    High priority stability enhancements

    • Fixed an error handling issue related to out-of-memory conditions which could result in brief periods of unavailability.

    • Fixed an issue when the connection to SQL Server fails using the TDS_FDW extension to query a foreign table.

    • Fixed an issue that caused connections using the provided root certificate to fail.

    • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

    Aurora PostgreSQL 11.16.0, June 9, 2022

    New features

    Critical updates

    • Fixed a replay crash due to an LSN mismatch.

    • Fixed the S3 extension to prevent invalid region injection.

    High stability updates

    • Fixed multiple issues related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an Aurora Serverless v2 scaling issue.

    General stability updates

    • Fixed a lock contention crash during an Aurora Serverless v1 scaling event.

    • Fixed an issue where logical replication becomes stuck after a restart.

    • Fixed multiple issues that could lead to brief periods of unavailability.

    • Fixed a crash in pg_cron due to a task still running but being unscheduled.

    • Fixed, during redo, an invalid page hit on the Generic Redo for GENERIC_XLOG_FULL_PAGE_DATA. This happens due to a timing hole between generating the log record and then writing the metadata for the record on the RW node and the RO node replays between those operations.

    • Improved the query performance by supporting parallel workers.

    • Upgraded the plugin wal2json version to 2.4.

    • Upgraded the pglogical extension to version 2.4.1.

    PostgreSQL 11.15

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.15. For more information about the improvements in PostgreSQL 11.15, see PostgreSQL release 11.15.

    Aurora PostgreSQL 11.15.4, July 18, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Critical enhancements

    • Fixed an issue during a storage node restart that could result in periods of unavailability.

    High stability enhancements

    • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

    • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

    • Fixed a defect where the validation of cached plans may lead to a database restart when the plan was previously invalidated.

    Aurora PostgreSQL 11.15.1, April 27, 2022

    High priority stability enhancements

    • Fixed an issue that could cause incorrect WriteIOPS reporting in the Amazon console.

    • Fixed an issue that could cause unavailability after removal of a read node from a cluster.

    General enhancements

    • Fixed an issue that could cause an engine restart during periods of low free memory.

    Aurora PostgreSQL 11.15.0, March 29, 2022

    High priority stability enhancements

    • Fixed multiple issues that may result in unavailability of a read node.

    • Fixed an issue that may result in a read node being unable to replay WAL requiring the replication slot to be dropped and resynchronized.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    General enhancements

    • Fixed a small memory leak on read nodes when commit_ts is set.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    • Fixed an issue that could cause an import from S3 to fail when using the S3 Extension.

    • Fixed multiple issues that could result in periods of unavailability when using apg_plan_mgmt.

    • Fixed multiple issues that could result in periods of unavailability when QPM is enabled.

    PostgreSQL 11.14

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.14. For more information about the improvements in PostgreSQL 11.14, see PostgreSQL release 11.14.

    Aurora PostgreSQL 11.14.4, July 20, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Critical enhancements

    • Fixed an issue during a storage node restart that could result in periods of unavailability.

    High stability enhancements

    • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

    • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v2.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    Aurora PostgreSQL 11.14.3, April 13, 2022

    General enhancements

    • Fixed a bug that could cause an engine restart during periods of low free memory.

    Aurora PostgreSQL 11.14.1

    Security enhancements

    • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 11.14.0

    Critical stability enhancements

    • Fixed a bug where logical replication may hang resulting in the replay falling behind on the read node. The instance may eventually restart.

    Additional improvements and enhancements

    • Fixed a buffer cache bug that could cause brief periods of unavailability.

    • Fixed a bug in the apg_plan_mgmt extension where an index based plan was not being enforced.

    • Fixed a bug in the pg_logical extension that could cause brief periods of unavailability due to improper handling of NULL arguments.

    • Fixed an issue where orphaned files caused major version upgrades to fail.

    • Fixed incorrect Aurora Storage Daemon log write metrics.

    • Fixed multiple bugs that could result in WAL replay falling behind and eventually causing the reader instances to restart.

    • Improved the Aurora buffer cache page validation on reads.

    • Improved the Aurora storage metadata validation.

    • Updated the pg_hint_pan extension to v1.3.7.

    PostgreSQL 11.13

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.13. For more information about the improvements in PostgreSQL 11.13, see PostgreSQL release 11.13.

    Aurora PostgreSQL 11.13.4, July 6, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    General enhancements

    • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that could cause a restart of the postmaster process in Amazon Aurora Serverless v2.

    • Fixed an issue that could cause Performance Insights to display "Unknown wait event".

    • Fixed an issue that could result in periods of unavailability due to the existence of duplicate relation files.

    Aurora PostgreSQL 11.13.3, June 6, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    High priority stability updates

    • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v1.

    • Fixed an issue that can cause a restart of the Aurora Runtime process in Amazon Aurora Serverless v1.

    General enhancements

    • Fixed a memory leak in the Aurora Runtime that could lead to an out-of-memory condition.

    Aurora PostgreSQL 11.13.2, April 12, 2022

    General enhancements

    • Fixed a buffer cache bug that could cause brief periods of unavailability.

    Aurora PostgreSQL 11.13.1

    Security enhancements

    • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 11.13.0

    Critical stability enhancements

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability enhancements

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue where read queries may time out on read nodes during the replay of lazy truncation triggered by vacuum on the write node.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue where the aurora_postgres_replica_status() function returned stale or lagging CPU stats.

    • Fixed an issue where, in rare cases, an Aurora Global Database secondary mirror cluster may restart due to a stall in the log apply process.

    • Fixed an issue with the apg_plan_mgmt extension where the planning and execution times were reported as 0.

    • Removed support for DES, 3DES and RC4 cipher suites.

    • Updated PostGIS extension to version 3.1.4.

    • Added support for postgis_raster extension version 3.1.4.

    PostgreSQL 11.12, Aurora PostgreSQL 3.6

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.12. For more information about the improvements in PostgreSQL 11.12, see PostgreSQL release 11.12.

    Aurora PostgreSQL 11.12.4, July 14, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    High priority stability enhancements

    • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    Aurora PostgreSQL 3.6.2

    Security enhancements

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched the PostgreSQL community fix for CVE-2021-3677: “Memory disclosure in certain queries”. CVE-2020-14350

    • Backpatched postgis to PostGIS 2.5.2. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 3.6.1

    Critical stability enhancements

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability enhancements

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    • Fixed an out of memory crash issue with Aurora storage daemon that leads to writer node restart. This also reduces the overall system memory consumption.

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue with the apg_plan_mgmt extension where planning and execution time were reported as 0.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue where in rare cases, an Aurora Global Database secondary mirror cluster may restart due to a stall in the log apply process.

    • Fixed an issue where orphaned files caused failed translations in read codepaths during or after major version upgrade.

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    • Fixed an out of memory crash issue with Aurora storage daemon that leads to writer node restart. This also reduces the overall system memory consumption.

    Aurora PostgreSQL 3.6.0

    High priority stability enhancements

    • Fixed an issue where creating a database from an existing template database with tablespace resulted in an error with the message ERROR: could not open file pg_tblspc/...: No such file or directory.

    • Fixed an issue where, in rare cases, an Aurora replica may be unable to start when a large number of PostgreSQL subtransactions (i.e. SQL savepoints) have been used.

    • Fixed an issue where, in rare circumstances, read results may be inconsistent for repeated read requests on replica nodes.

    Additional improvements and enhancements

    • Upgraded OpenSSL to 1.1.1k.

    • Reduced CPU usage and memory consumption of the WAL apply process on Aurora replicas for some workloads.

    • Improved metadata protection from accidental erasure.

    • Improved safety checks in the write path to detect incorrect writes to metadata.

    • Improved security by removing 3DES and other older ciphers for SSL/TLS connections.

    • Fixed an issue where a duplicate file entry can prevent the Aurora PostgreSQL engine from starting up.

    • Fixed an issue that could cause temporary unavailability under heavy workloads.

    • Added back ability to use a leading forward slash in the S3 path during S3 import.

    • Updated the orafce extension to version 3.16.

    PostgreSQL 11.11, Aurora PostgreSQL 3.5 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.11. For more information about the improvements in PostgreSQL 11.11, see PostgreSQL release 11.11.

    Aurora PostgreSQL 3.5.1

    Security enhancements

    • Modified the pg_bigm extension to mitigate a security issue during create extension. The issue was addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched the PostgreSQL community fix for CVE-2021-3677: “Memory disclosure in certain queries”. For more information, see CVE-2021-3677

    • Backpatched postgis to PostGIS 2.5.2. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched an input validation error in the log_fdw extension function parameters.

    Aurora PostgreSQL 3.5.0

    New features

    • Added support for the following extensions:

      • The pg_proctab extension version 0.0.9

      • The pg_bigm extension version 1.2

    High priority stability enhancements

    • Fixed a bug where in rare cases a reader had inconsistent results when it restarted while a transaction with more than 64 subtransactions was being processed.

    • Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    • Fixed a bug where the database could not be started when there were many relations in memory-constrained environments.

    • Fixed a bug in the apg_plan_mgmt extension that could cause brief periods of unavailability due to an internal buffer overflow.

    • Fixed a bug on reader nodes that could cause brief periods of unavailability during WAL replay.

    • Fixed a bug in the rds_activity_stream extension that caused an error during startup when attempting to log audit events.

    • Fixed bugs in the aurora_replica_status function where rows were sometimes partially populated and some values such as Replay Latency, and CPU usage were always 0.

    • Fixed a bug where the database engine would attempt to create shared memory segments larger than the instance total memory and fail repeatedly. For example, attempts to create 128 GiB shared buffers on a db.r5.large instance would fail. With this change, requests for total shared memory allocations larger than the instance memory allow setting the instance to incompatible parameters.

    • Added logic to clean up unnecessary pg_wal temporary files on a database startup.

    • Fixed a bug that reported ERROR: rds_activity_stream stack item 2 not found on top - cannot pop when attempting to create the rds_activity_stream extension.

    • Fixed a bug that could cause the error failed to build any 3-way joins in a correlated IN subquery under an EXISTS subquery.

    • Backported the following performance improvement from the PostgreSQL community: pg_stat_statements: add missing check for pgss_enabled().

    • Fixed a bug that could cause brief periods of unavailability due to running out of memory when creating the postgis extension with pgAudit enabled.

    • Fixed a bug when using outbound logical replication to synchronize changes to another database that could fail with an error message like ERROR: could not map filenode "base/16395/228486645" to relation OID.

    • Fixed a bug that could cause a brief period of unavailability when aborting a transaction.

    • Fixed a bug where the rds_ad role wasn't created after upgrading from a version of Aurora PostgreSQL that doesn't support Microsoft Active Directory authentication.

    • Added btree page checks to detect tuple metadata inconsistency.

    • Fixed a bug in asynchronous buffer reads that could cause brief periods of unavailability on reader nodes during WAL replay.

    PostgreSQL 11.9, Aurora PostgreSQL 3.4

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.9. For more information about the improvements in PostgreSQL 11.9, see PostgreSQL release 11.9.

    Aurora PostgreSQL 11.9.6, July 8, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Aurora PostgreSQL 3.4.5

    Security enhancements

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched the PostgreSQL community fix for CVE-2021-3677: “Memory disclosure in certain queries”. For more information, see CVE-2021-3677

    • Backpatched the PostgreSQL community fix for CVE-2021-3393: "Partition constraint violation errors leak values of denied columns". For more information, see CVE-2021-3393

    • Backpatched postgis to PostGIS 2.5.2. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched an input validation error in the log_fdw extension function parameters.

    Aurora PostgreSQL 3.4.3

    High priority stability enhancements

    • Provided a patch for PostgreSQL community security issues CVE-2021-32027, CVE-2021-32028 and CVE-2021-32029.

    Additional improvements and enhancements

    • Fixed a bug in the aws_s3 extension to allow import of objects with leading forward slashes in the object identifier.

    • Fixed a bug in the rds_activity_stream extension that caused an error during startup when attempting to log audit events.

    • Fixed a bug that returned an ERROR when attempting to create the rds_activity_stream extension.

    • Fixed a bug that could cause brief periods of unavailability due to running out of memory when creating the postgis extension with pgAudit enabled.

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    Aurora PostgreSQL 3.4.2

    High priority stability enhancements

    • Fixed a bug where in rare cases a reader had inconsistent results when it restarted while a transaction with more than 64 subtransactions was being processed.

    Additional improvements and enhancements

    • Fixed a bug that could lead to intermittent unavailability due to a race condition when handling responses from storage nodes.

    • Fixed a bug that could lead to intermittent unavailability due to the rotation of network encryption keys.

    • Fixed a bug that could lead to intermittent unavailability due to heat management of the underlying storage segments.

    • Fixed a bug where a large S3 import with thousands of clients can cause one or more of the import clients to stop responding.

    • Removed a restriction that prevented setting configuration variable strings that contained brazil.

    • Fixed a bug that could lead to intermittent unavailability if a reader node runs a query that access many tables while the writer node is acquiring exclusive locks on all of the same tables.

    Aurora PostgreSQL 3.4.1

    Critical stability enhancements

    • Fixed a bug that caused a read replica to unsuccessfully restart repeatedly in rare cases.

    • Fixed a bug where a cluster became unavailable when attempting to create more than 16 read replicas or Aurora global database secondary Amazon Web Services Regions. The cluster became available again when the new read replica or secondary Amazon Web Services Region was removed.

    Additional improvements and enhancements

    • Fixed a bug that when under heavy load, snapshot import, COPY import, or S3 import stopped responding in rare cases.

    • Fixed a bug where a read replica might not join the cluster when the writer was very busy with a write-intensive workload.

    • Fixed a bug where a cluster could be unavailable briefly when a high-volume S3 import was running.

    • Fixed a bug that caused a cluster to take several minutes to restart if a logical replication stream was terminated while handling many complex transactions.

    • Disallowed the use of both IAM and Kerberos authentication for the same user.

    Aurora PostgreSQL 3.4.0

    New features

    Critical stability enhancements

    • None

    High priority stability enhancements

    • Fixed a bug in Aurora PostgreSQL replication that could result in the error message ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    • Fixed a bug where in some cases, DB clusters that have logical replication enabled did not remove truncated WAL segment files from storage. This resulted in volume size growth.

    • Backported fixes for the following PostgreSQL community security issues:

    • Fixed a bug in the pg_stat_statements extension that caused excessive CPU consumption.

    Additional improvements and enhancements

    • You can now use pg_replication_slot_advance to advance a logical replication slot for the roles rds_replication and rds_superuser.

    • Improved the asynchronous mode performance of database activity streams.

    • Reduced the delay when publishing to CloudWatch the rpo_lag_in_msec metric for Aurora global database clusters.

    • Aurora PostgreSQL no longer falls behind on a read node when the backend is blocked writing to the database client.

    • Fixed a bug that in rare cases caused a brief period of unavailability on a read replica when the storage volume grew.

    • Fixed a bug when creating a database that could return the following: ERROR: could not create directory on local disk

    • Updated data grid files to fix errors or incorrect transformation results from the ST_Transform method of the PostGIS extension.

    • Fixed a bug where in some cases replaying XLOG_BTREE_REUSE_PAGE records on Aurora reader instances caused unnecessary replay lag.

    • Fixed a small memory leak in a b-tree index that could lead to an out of memory condition.

    • Fixed a bug in the GiST index that could result in an out of memory condition after promoting an Aurora read replica.

    • Fixed an S3 import bug that reported ERROR: HTTP 403. Permission denied when importing data from a file inside an S3 subfolder.

    • Fixed a bug in the aws_s3 extension for pre-signed URL handling that could result in the error message S3 bucket names with a period (.) are not supported.

    • Fixed a bug in the aws_s3 extension where an import might be blocked indefinitely if an exclusive lock was taken on the relation prior to beginning the operation.

    • Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance that uses GiST indexes. In rare cases, this bug caused a brief period of unavailability after promoting the Aurora cluster.

    • Fixed a bug in database activity streams where customers were not notified of the end of an outage.

    • Updated the pg_audit extension to version 1.3.1.

    PostgreSQL 11.8, Aurora PostgreSQL 3.3 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.8. For more information about the improvements in PostgreSQL 11.8, see PostgreSQL release 11.8.

    Aurora PostgreSQL release 3.3.2

    Critical stability enhancements

    • None

    High priority stability enhancements

    • Fixed a bug in Aurora PostgreSQL replication that could result in the error message ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    • Fixed a bug where in some cases, DB clusters that have logical replication enabled did not remove truncated WAL segment files from storage. This resulted in volume size growth.

    • Fixed an issue with creating a global database cluster in a secondary region.

    • Backported fixes for the following PostgreSQL community security issues:

    • Fixed a bug in the pg_stat_statements extension that caused excessive CPU consumption.

    Additional improvements and enhancements

    • Aurora PostgreSQL no longer falls behind on a read node when the backend is blocked writing to the database client.

    • Reduced the delay when publishing to CloudWatch the rpo_lag_in_msec metric for Aurora global database clusters.

    • Fixed a bug where a DROP DATABASE statement didn't remove any relation files.

    • Fixed a bug where in some cases replaying XLOG_BTREE_REUSE_PAGE records on Aurora reader instances caused unnecessary replay lag.

    • Fixed a small memory leak in a b-tree index that could lead to an out of memory condition.

    • Fixed a bug in the aurora_replica_status() function where the server_id field was sometimes truncated.

    • Fixed a bug where a log record was incorrectly processed causing the Aurora replica to crash.

    • Fixed an S3 import bug that reported ERROR: HTTP 403. Permission denied when importing data from a file inside an S3 subfolder.

    • You can now use pg_replication_slot_advance to advance a logical replication slot for the roles rds_replication and rds_superuser.

    • Improved performance of the asynchronous mode for database activity streams.

    • Fixed a bug in the aws_s3 extension that could result in the error message S3 bucket names with a period (.) are not supported.

    • Fixed a race condition that caused valid imports to intermittently fail.

    • Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance that uses GiST indexes. In rare cases, this bug caused a brief period of unavailability after promoting the Aurora DB cluster.

    • Fixed a bug in the aws_s3 extension where an import may be blocked indefinitely if an exclusive lock was taken on the relation prior to beginning the operation.

    Aurora PostgreSQL 3.3.1

    You can find the following improvements in this release.

    Critical stability enhancements

    1. Fixed a bug that appears when the NOT EXISTS operator incorrectly returns TRUE, which can only happen when the following unusual set of circumstances occurs:

      • A query is using the NOT EXISTS operator.

      • The column or columns being evaluated against the outer query in the NOT EXISTS subquery contain a NULL value.

      • There isn't a another predicate in the subquery that removes the need for the evaluation of the NULL values.

      • The filter used in the subquery does not use an index seek for its execution.

      • The operator isn't converted to a join by the query optimizer.

    Aurora PostgreSQL 3.3.0

    New features

    • Added support for the RDKit extension version 3.8.

      The RDKit extension provides modeling functions for cheminformatics. Cheminformatics is storing, indexing, searching, retrieving, and applying information about chemical compounds. For example, with the RDKit extension you can construct models of molecules, search for molecular structures, and read or create molecules in various notations. You can also perform research on data loaded from the ChEMBL website or a SMILES file. The Simplified Molecular Input Line Entry System (SMILES) is a typographical notation for representing molecules and reactions. For more information, see The RDKit database cartridge in the RDKit documentation.

    • Added support for a minimum TLS version

      Support for a minimum Transport Layer Security (TLS) version is back ported from PostgreSQL 12. It allows the Aurora PostgreSQL server to constrain the TLS protocols with which a client is allowed to connect via two new PostgreSQL parameters. These parameters include ssl_min_protocol_versionand ssl_max_protocol_version. For example, to limit client connections to the Aurora PostgreSQL server to at least TLS 1.2 protocol version, set the ssl_min_protocol_version to TLSv1.2.

    • Added support for the pglogical extension version 2.2.2.

      The pglogical extension is a logical streaming replication system that provides additional features beyond what's available in PostgreSQL native logical replication. Features include conflict handling, row filtering, DDL/sequence replication and delayed apply. You can use the pglogical extension to set up replication between Aurora PostgreSQL clusters, between RDS PostgreSQL and Aurora PostgreSQL, and with PostgreSQL databases running outside of RDS.

    • Aurora dynamically resizes your cluster storage space. With dynamic resizing, the storage space for your Aurora DB cluster automatically decreases when you remove data from the DB cluster. For more information, see Storage scaling in the Amazon Aurora User Guide.

      Note

      The dynamic resizing feature is being deployed in phases to the Amazon Web Services Regions where Aurora is available. Depending on the Region where your cluster is, this feature might not be available yet. For more information, see the What's New announcement.

    Critical stability enhancements

    • Fixed a bug related to heap page extend that in rare cases resulted in longer recovery time and impacted availability.

    High priority stability enhancements

    • Fixed a bug in Aurora Global Database that could cause delays in upgrading the database engine in a secondary Amazon Web Services Region. For more information, see Using Amazon Aurora global databases in the Amazon Aurora User Guide.

    • Fixed a bug that in rare cases caused delays in upgrading a database to engine version 11.8.

    Additional improvements and enhancements

    • Fixed a bug where the Aurora replica crashed when workloads with heavy subtransactions are made on the writer instance.

    • Fixed a bug where the writer instance crashed due to a memory leak and the depletion of memory used to track active transactions.

    • Fixed a bug that lead to a crash due to improper initialization when there is no free memory available during PostgreSQL backend startup.

    • Fixed a bug where an Aurora PostgreSQL Serverless DB cluster might return the following error after a scaling event: ERROR: prepared statement "S_6" already exists.

    • Fixed an out-of-memory problem when issuing the CREATE EXTENSION command with PostGIS when Database Activity Streams enabled.

    • Fixed a bug where a SELECT query might incorrectly return the error Attempting to read past EOF of relation rrrr. blockno=bbb nblocks=nnn.

    • Fixed a bug where the database might be unavailable briefly due to error handling in database storage growth.

    • Fixed a bug in Aurora PostgreSQL Serverless where queries that executed on previously idle connections got delayed until the scale operation completed.

    • Fixed a bug where an Aurora PostgreSQL DB cluster with Database Activity Streams enabled might report the beginning of a potential loss window for activity records, but does not report the restoration of connectivity.

    • Fixed a bug with the aws_s3.table_import_from_s3 function where a COPY from S3 failed with the error HTTP error code: 248. For more information, see aws_s3.table_import_from_s3 in the Amazon Aurora User Guide.

    PostgreSQL 11.7, Aurora PostgreSQL 3.2 (Deprecated on March 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.7. For more information about the improvements in PostgreSQL 11.7, see PostgreSQL release 11.7.

    Aurora PostgreSQL 3.2.7

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 3.2.6

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    • Fixed a bug in Aurora PostgreSQL replication that might result in the error message, ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    Additional improvements and enhancements

    • Fixed a bug that in rare cases caused brief read replica unavailability when storage volume grew.

    • Aurora PostgreSQL Serverless now supports execution of queries on all connections during a scale event.

    • Fixed a bug in Aurora PostgreSQL Serverless where a leaked lock resulted in a prolonged scale event.

    • Fixed a bug where the aurora_replica_status function showed truncated server identifiers.

    • Fixed a bug in Aurora PostgreSQL Serverless where connections being migrated during a scale event disconnected with the message: "ERROR: could not open relation with OID ....

    • Fixed a small memory leak in a b-tree index that could lead to an out of memory condition.

    • Fixed a bug in a GiST index that might result in an out of memory condition after promoting an Aurora Read Replica.

    • Improved performance for Database Activity Streams.

    • Fixed a bug in Database Activity Streams where customers were not notified when an outage ended.

    • Fixed a bug in the aws_s3 extension for pre-signed URL handling that could have resulted in the error message S3 bucket names with a period (.) are not supported.

    • Fixed a bug in the aws_s3 extension where incorrect error handling could lead to failures during the import process.

    • Fixed a bug in the aws_s3 extension where an import may be blocked indefinitely if an exclusive lock was taken on the relation prior to beginning the operation.

    Aurora PostgreSQL 3.2.4

    You can find the following improvements in this release.

    Critical stability enhancements

    1. Fixed a bug that appears when the NOT EXISTS operator incorrectly returns TRUE, which can only happen when the following unusual set of circumstances occurs:

      • A query is using the NOT EXISTS operator.

      • The column or columns being evaluated against the outer query in the NOT EXISTS subquery contain a NULL value.

      • There isn't a another predicate in the subquery that removes the need for the evaluation of the NULL values.

      • The filter used in the subquery does not use an index seek for its execution.

      • The operator isn't converted to a join by the query optimizer.

    Aurora PostgreSQL 3.2.3

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    • None

    Additional improvements and enhancements

    • Fixed a bug in Aurora PostgreSQL Serverless where queries that ran on previously idle connections got delayed until the scale operation completed.

    • Fixed a bug that might cause brief unavailability for heavy subtransaction workloads when multiple reader instances restart or rejoin the cluster.

    Aurora PostgreSQL 3.2.2

    You can find the following improvements in this release.

    Critical stability enhancements

    • Fixed a bug related to heap page extend that in rare cases resulted in longer recovery time and impacted availability.

    High priority stability enhancements

    • Fixed a bug in Aurora Global Database that could cause delays in upgrading the database engine in a secondary region. For more information, see Using Amazon Aurora global databases in the Amazon Aurora User Guide.

    • Fixed a bug that in rare cases caused delays in upgrading a database to engine version 11.7.

    Additional improvements and enhancements

    • Fixed a bug where the database might be unavailable briefly due to error handling in database storage growth.

    • Fixed a bug where a SELECT query might incorrectly return the error, Attempting to read past EOF of relation rrrr. blockno=bbb nblocks=nnn.

    • Fixed a bug where an Aurora PostgreSQL Serverless DB cluster might return the following error after a scaling event: ERROR: prepared statement "S_6" already exists.

    Aurora PostgreSQL 3.2.1

    New features

    You can find the following improvements in this release.

    Critical stability enhancements

    None.

    High priority stability enhancements

    • Improved performance and availability of read instances when applying DROP TABLE and TRUNCATE TABLE operations.

    • Fixed a small but continuous memory leak in a diagnostic module that could lead to an out-of-memory condition on smaller DB instance types.

    • Fixed a bug in the PostGIS extension which could lead to a database restart. This has been reported to the PostGIS community as https://trac.osgeo.org/postgis/ticket/4646.

    • Fixed a bug where read requests might stop responding due to incorrect error handling in the storage engine.

    • Fixed a bug that fails for some queries and results in the message ERROR: found xmin xxxxxx from before relfrozenxid yyyyyyy. This could occur following the promotion of a read instance to a write instance.

    • Fixed a bug where an Aurora serverless DB cluster might crash while rolling back a scale attempt.

    Additional improvements and enhancements

    • Improved performance for queries that read many rows from storage.

    • Improved performance and availability of reader DB instances during heavy read workload.

    • Enabled correlated IN and NOT IN subqueries to be transformed to joins when possible.

    • Improved the filtering estimation for enhanced semi-join filter pushdown by using multi-column statistics or indexes when available.

    • Improved read performance of the pg_prewarm extension.

    • Fixed a bug where an Aurora serverless DB cluster might report the message ERROR: incorrect binary data format in bind parameter ... following a scale event.

    • Fixed a bug where a serverless DB cluster might report the message ERROR: insufficient data left in message following a scale event.

    • Fixed a bug where an Aurora serverless DB cluster can experience prolonged or failed scale attempts.

    • Fixed a bug that resulted in the message ERROR: could not create file "base/xxxxxx/yyyyyyy" as a previous version still exists on disk: Success. Please contact Amazon customer support. This can occur during object creation after PostgreSQL's 32-bit object identifier has wrapped around.

    • Fixed a bug where the write-ahead-log (WAL) segment files for PostgreSQL logical replication were not deleted when changing the wal_level value from logical to replica.

    • Fixed a bug in the pg_hint_plan extension where a multi-statement query could lead to a crash when enable_hint_table is enabled. This is tracked in the PostgreSQL community as https://github.com/ossc-db/pg_hint_plan/issues/25.

    • Fixed a bug where JDBC clients might report the message java.io.IOException: Unexpected packet type: 75 following a scale event in an Aurora serverless DB cluster.

    • Fixed a bug in PostgreSQL logical replication that resulted in the message ERROR: snapshot reference is not owned by resource owner TopTransaction.

    • Changed the following extensions:

      • Updated orafce to version 3.8

      • Updated pgTAP to version 1.1

    • Provided support for fault injection queries.

    PostgreSQL 11.6, Aurora PostgreSQL 3.1 (Deprecated on March 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.6. For more information about the improvements in PostgreSQL 11.6, see PostgreSQL release 11.6.

    This release contains multiple critical stability enhancements. Amazon highly recommends upgrading your Aurora PostgreSQL clusters that use older PostgreSQL 11 engines to this release.

    Aurora PostgreSQL 3.1.4

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 3.1.3

    New features

    • Aurora PostgreSQL now supports the PostgreSQL vacuum_truncate storage parameter to manage vacuum truncation for specific tables. Set this storage parameter to false for a table to prevent the VACUUM SQL command from truncating the table's trailing empty pages.

    Critical stability enhancements

    • None

    High priority stability enhancements

    • Fixed a bug where reads from storage might stop responding due to incorrect error handling.

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 3.1.2

    This release contains a critical stability enhancement. Amazon highly recommends updating your older Aurora PostgreSQL 11-compatible clusters to this release.

    Critical stability enhancements

    • Fixed a bug in which a reader DB instance might temporarily use stale data. This could lead to wrong results such as too few or too many rows. This error is not persisted on storage, and will clear when the database page containing the row has been evicted from cache. This can happen when the primary DB instance enters a transaction snapshot overflow due to having more than 64 subtransactions in a single transaction. Applications susceptible to this bug include those that use SQL savepoints or PostgreSQL exception handlers with more than 64 subtransactions in the top transaction.

    High priority stability enhancements

    • Fixed a bug that might cause a reader DB instance to crash causing unavailability while attempting to the join the DB cluster. This can happen in some cases when the primary DB instance has a transaction snapshot overflow due to a high number of subtransactions. In this situation the reader DB instance will be unable to join until the snapshot overflow has cleared.

    Additional improvements and enhancements

    • Fixed a bug that prevented Performance Insights from determining the query ID of a running statement.

    Aurora PostgreSQL 3.1.1

    You can find the following improvements in this release.

    Critical stability enhancements

    • Fixed a bug in which the DB instance might be briefly unavailable due to the self-healing function of the underlying storage.

    High priority stability enhancements

    • Fixed a bug in which the database engine might crash causing unavailability. This occurred while scanning an included, non-key column of a B-Tree index. This only applies to PostgreSQL 11 "included column" indexes.

    • Fixed a bug that might cause the database engine to crash causing unavailability. This occurred if a newly established database connection encountered a resource exhaustion-related error during initialization after successful authentication.

    Additional improvements and enhancements

    • Provided a fix for the pg_hint_plan extension that could lead the database engine to crash causing unavailability. The open source issue can be tracked at https://github.com/ossc-db/pg_hint_plan/pull/45.

    • Fixed a bug where SQL of the form ALTER FUNCTION ... OWNER TO ... incorrectly reported ERROR: improper qualified name (too many dotted names).

    • Improved the performance of GIN index vacuum via prefetching.

    • Fixed a bug in open source PostgreSQL that could lead to a database engine crash causing unavailability. This occurred during parallel B-Tree index scans. This issue has been reported to the PostgreSQL community.

    • Improved the performance of in-memory B-Tree index scans.

    For information about extensions and modules, see Extensions supported for Aurora PostgreSQL 11.

    Aurora PostgreSQL 3.1.0

    You can find the following new features and improvements in this engine version.

    New features

    1. Support for exporting data to Amazon S3. For more information, see Exporting data from an Aurora PostgreSQL DB cluster to Amazon S3 in the Amazon Aurora User Guide.

    2. Support for Amazon Aurora Machine Learning. For more information, see Using machine learning (ML) with Aurora PostgreSQL in the Amazon Aurora User Guide.

    3. SQL processing enhancements include:

      • Optimizations for NOT IN with the apg_enable_not_in_transform parameter.

      • Semi-join filter pushdown enhancements for hash joins with the apg_enable_semijoin_push_down parameter.

      • Optimizations for redundant inner join removal with the apg_enable_remove_redundant_inner_joins parameter.

      • Improved ANSI compatibility options with the ansi_constraint_trigger_ordering, ansi_force_foreign_key_checks and ansi_qualified_update_set_target parameters.

      For more information, see Amazon Aurora PostgreSQL parameters in the Amazon Aurora User Guide.

    4. New and updated PostgreSQL extensions include:

    Critical stability enhancements

    1. Fixed a bug related to creating B-tree indexes on temporary tables that in rare cases might result in longer recovery time, and impact availability.

    2. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance. In rare cases, this bug causes a log write failure that might result in longer recovery time, and impact availability.

    3. Fixed a bug related to handling of reads with high I/O latency that in rare cases might result in longer recovery time, and impact availability.

    High priority stability enhancements

    1. Fixed a bug related to logical replication in which wal segments are not properly removed from storage. This can result in storage bloat. To monitor this, view the TransactionLogDiskUsage parameter.

    2. Fixed multiple bugs, which cause Aurora to crash during prefetch operations on Btree indexes.

    3. Fixed a bug in which an Aurora restart might time out when logical replication is used.

    4. Enhanced the validation checks performed on data blocks in the buffer cache. This improves Aurora's detection of inconsistency.

    Additional improvements and enhancements

    1. The query plan management extension apg_plan_mgmt has an improved algorithm for managing plan generation for highly partitioned tables.

    2. Reduced startup time on instances with large caches via improvements in the buffer cache recovery algorithm.

    3. Improved the performance of the read-node-apply process under high transaction rate workloads by using changes to PostgreSQL LWLock prioritization. These changes prevent starvation of the read-node-apply process while the PostgreSQL ProcArray is under heavy contention.

    4. Improved handling of batch reads during vacuum, table scans, and index scans. This results in greater throughput and lower CPU consumption.

    5. Fixed a bug in which a read node might crash during the replay of a PostgreSQL SLRU-truncate operation.

    6. Fixed a bug where in rare cases, database writes might stall following an error returned by one of the six copies of an Aurora log record.

    7. Fixed a bug related to logical replication where an individual transaction larger than 1 GB in size might result in an engine crash.

    8. Fixed a memory leak on read nodes when cluster cache management is enabled.

    9. Fixed a bug in which importing an RDS PostgreSQL snapshot might stop responding if the source snapshot contains a large number of unlogged relations.

    10. Fixed a bug in which the Aurora storage daemon might crash under heavy I/O load.

    11. Fixed a bug related to hot_standby_feedback for read nodes in which the read node might report the wrong transaction id epoch to the write node. This can cause the write node to ignore the hot_standby_feedback and invalidate snapshots on the read node.

    12. Fixed a bug in which storage errors that occur during CREATE DATABASE statements are not properly handled. The bug left the resulting database inaccessible. The correct behavior is to fail the database creation and return the appropriate error to the user.

    13. Improved handling of PostgreSQL snapshot overflow when a read node attempts to connect to a write node. Prior to this change, if the write node was in a snapshot overflow state, the read node was unable to join. A message appeared in the PostgreSQL log file in the form DEBUG: recovery snapshot waiting for non-overflowed snapshot or until oldest active xid on standby is at least xxxxxxx (now yyyyyyy). A snapshot overflow occurs when an individual transaction has created over 64 subtransactions.

    14. Fixed a bug related to common table expressions in which an error is incorrectly raised when a NOT IN class exists in a CTE. The error is CTE with NOT IN fails with ERROR: could not find CTE CTE-Name.

    15. Fixed a bug related to an incorrect last_error_timestamp value in the aurora_replica_status table.

    16. Fixed a bug to avoid populating shared buffers with blocks belonging to temporary objects. These blocks correctly reside in PostgreSQL backend local buffers.

    17. Changed the following extensions:

      • Updated pg_hint_plan to version 1.3.4.

      • Added plprofiler version 4.1.

      • Added pgTAP version 1.0.0.

    PostgreSQL 11.4, Aurora PostgreSQL 3.0 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 11.4 with the Aurora PostgreSQL 3.0 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 11.4. For more information about the improvements in PostgreSQL 11.4, see PostgreSQL release 11.4.

    You can find the following improvements in this release.

    Improvements

    1. This release contains all fixes, features, and improvements present in Aurora PostgreSQL 2.3.5.

    2. Partitioning – Partitioning improvements include support for hash partitioning, enabling creation of a default partition, and dynamic row movement to another partition based on the key column update.

    3. Performance – Performance improvements include parallelism while creating indexes, materialized views, hash joins, and sequential scans to make the operations perform better.

    4. Stored procedures – SQL stored procedures now added support for embedded transactions.

    5. Autovacuum improvements – To provide valuable logging, the parameter rds.force_autovacuum_logging is ON by default in conjunction with the log_autovacuum_min_duration parameter set to 10 seconds. To increase autovacuum effectiveness, the values for the autovacuum_max_workers and autovacuum_vacuum_cost_limit parameters are computed based on host memory capacity to provide larger default values.

    6. Improved transaction timeout – The parameter idle_in_transaction_session_timeout is set to 24 hours. Any session that has been idle more than 24 hours is terminated.

    7. The tsearch2 module is no longer supported – If your application uses tsearch2 functions, update it to use the equivalent functions provided by the core PostgreSQL engine. For more information about the tsearch2 module, see PostgreSQL tsearch2.

    8. The chkpass module is no longer supported – For more information about the chkpass module, see PostgreSQL chkpass.

    9. Updated the following extensions:

      • address_standardizer to version 2.5.1

      • address_standardizer_data_us to version 2.5.1

      • btree_gin to version 1.3

      • citext to version 1.5

      • cube to version 1.4

      • hstore to version 1.5

      • ip4r to version 2.2

      • isn to version 1.2

      • orafce to version 3.7

      • pg_hint_plan to version 1.3.4

      • pg_prewarm to version 1.2

      • pg_repack to version 1.4.4

      • pg_trgm to version 1.4

      • pgaudit to version 1.3

      • pgrouting to version 2.6.1

      • pgtap to version 1.0.0

      • plcoffee to version 2.3.8

      • plls to version 2.3.8

      • plv8 to version 2.3.8

      • postgis to version 2.5.1

      • postgis_tiger_geocoder to version 2.5.1

      • postgis_topology to version 2.5.1

      • rds_activity_stream to version 1.3

    PostgreSQL 10.21 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.21. For more information about the improvements in PostgreSQL 10.21, see PostgreSQL release 10.21.

    Aurora PostgreSQL 10.21.4, November 17, 2022

    High priority stability enhancements

    • Fixed an issue that can cause increased network traffic when a writer instance transmits logs to a replica instance.

    Aurora PostgreSQL 10.21.3, October 13, 2022

    High priority stability enhancements

    • Fixed a PLV8 issue where the base parameter doesn't get loaded properly into the memory.

    • Fixed an issue where scaling an Aurora Serverless v2 instance gets stuck if VACUUM is running.

    General stability enhancements

    • Fixed a bug where Aurora PostgreSQL can't file the relfilenode.

    • Fixed a stuck scaling issue when the current scaling event times out.

    • Upgraded the PostGIS extension to version 3.1.7.

    • Fixed an issue where extended query messages might be lost during zero-downtime patching (ZDP) causing the extended query to hang after the ZDP completion.

    Aurora PostgreSQL 10.21.1, July 6, 2022

    Critical stability enhancements

    • Fixed an issue that could cause periods of unavailability during a storage node restart.

    High priority stability enhancements

    • Fixed an error handling issue related to out-of-memory conditions which could result in brief periods of unavailability.

    • Fixed an issue when the connection to SQL Server fails using the TDS_FDW extension to query a foreign table.

    • Fixed an issue that caused connections using the provided root certificate to fail.

    • Improved the diagnostic and supportability information in case of inconsistent B-tree index entries.

    Aurora PostgreSQL 10.21.0, June 9, 2022

    New features

    Critical updates

    • Fixed a replay crash due to an LSN mismatch.

    • Fixed the S3 extension to prevent invalid region injection.

    High stability updates

    • Fixed multiple issues related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an Aurora Serverless v2 scaling issue.

    General stability updates

    • Fixed a lock contention crash during an Aurora Serverless v1 scaling event.

    • Fixed an issue where logical replication becomes stuck after a restart.

    • Fixed multiple issues that could lead to brief periods of unavailability.

    • Fixed a crash in pg_cron due to a task still running but being unscheduled.

    • Fixed, during redo, an invalid page hit on the Generic Redo for GENERIC_XLOG_FULL_PAGE_DATA. This happens due to a timing hole between generating the log record and then writing the metadata for the record on the RW node and the RO node replays between those operations.

    • Improved the query performance by supporting parallel workers.

    • Upgraded the plugin wal2json version to 2.4.

    • Upgraded the pglogical extension to version 2.4.1.

    PostgreSQL 10.20 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.20. For more information about the improvements in PostgreSQL 10.20, see PostgreSQL release 10.20.

    Aurora PostgreSQL 10.20.4, July 18, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Critical enhancements

    • Fixed an issue during a storage node restart that could result in periods of unavailability.

    High stability enhancements

    • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

    • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

    • Fixed a defect where the validation of cached plans may lead to a database restart when the plan was previously invalidated.

    Aurora PostgreSQL 10.20.1, April 27, 2022

    High priority stability enhancements

    • Fixed an issue that could cause incorrect WriteIOPS reporting in the Amazon console.

    • Fixed an issue that could cause unavailability after removal of a read node from a cluster.

    General enhancements

    • Fixed an issue that could cause an engine restart during periods of low free memory.

    Aurora PostgreSQL 10.20.0, March 29, 2022

    High priority stability enhancements

    • Fixed multiple issues that may result in unavailability of a read node.

    • Fixed an issue that may result in a read node being unable to replay WAL requiring the replication slot to be dropped and resynchronized.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    General enhancements

    • Fixed a small memory leak on read nodes when commit_ts is set.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    • Fixed an issue that could cause an import from S3 to fail when using the S3 Extension.

    • Fixed multiple issues that could result in periods of unavailability when using apg_plan_mgmt.

    • Fixed multiple issues that could result in periods of unavailability when QPM is enabled.

    PostgreSQL 10.19 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.19. For more information about the improvements in PostgreSQL 10.19, see PostgreSQL release 10.19.

    Aurora PostgreSQL 10.19.4, July 20, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    Critical enhancements

    • Fixed an issue during a storage node restart that could result in periods of unavailability.

    High stability enhancements

    • Fixed an error handling issue related to out of memory conditions that could result in brief periods of unavailability.

    • Fixed an issue related to the existence of duplicate relation files that could result in periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v2.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    Aurora PostgreSQL 10.19.3, April 13, 2022

    General enhancements

    • Fixed a bug that could cause an engine restart during periods of low free memory.

    Aurora PostgreSQL 10.19.1

    Security enhancements

    • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 10.19.0

    Critical stability enhancements

    • Fixed a bug where logical replication may hang resulting in the replay falling behind on the read node. The instance may eventually restart.

    Additional improvements and enhancements

    • Fixed a buffer cache bug that could cause brief periods of unavailability.

    • Fixed a bug in the apg_plan_mgmt extension where an index based plan was not being enforced.

    • Fixed a bug in the pg_logical extension that could cause brief periods of unavailability due to improper handling of NULL arguments.

    • Fixed an issue where orphaned files caused major version upgrades to fail.

    • Fixed incorrect Aurora Storage Daemon log write metrics.

    • Fixed multiple bugs that could result in WAL replay falling behind and eventually causing the reader instances to restart.

    • Improved the Aurora buffer cache page validation on reads.

    • Improved the Aurora storage metadata validation.

    • Updated the pg_hint_pan extension to v1.3.6.

    PostgreSQL 10.18 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.18. For more information about the improvements in PostgreSQL 10.18, see PostgreSQL release 10.18.

    Aurora PostgreSQL 10.18.4, July 6, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    General enhancements

    • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that could cause a restart of the postmaster process in Amazon Aurora Serverless v2.

    • Fixed an issue that could cause Performance Insights to display "Unknown wait event".

    • Fixed an issue that could result in periods of unavailability due to the existence of duplicate relation files.

    Aurora PostgreSQL 10.18.3, June 6, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    High priority stability updates

    • Fixed an issue that can cause a restart of the postmaster process in Amazon Aurora Serverless v1.

    • Fixed an issue that can cause a restart of the Aurora Runtime process in Amazon Aurora Serverless v1.

    General enhancements

    • Fixed a memory leak in the Aurora Runtime that could lead to an out-of-memory condition.

    Aurora PostgreSQL 10.18.2, April 12, 2022

    General updates

    • Fixed a buffer cache bug that could cause brief periods of unavailability.

    Aurora PostgreSQL 10.18.1

    Security enhancements

    • Updated the PostGIS extension from version 3.1.4 to 3.1.5. This update contains a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 10.18.0

    Critical stability updates

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability updates

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue where read queries may time out on read nodes during the replay of lazy truncation triggered by vacuum on the write node.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue where the aurora_postgres_replica_status() function returned stale or lagging CPU stats.

    • Fixed an issue where, in rare cases, an Aurora Global Database secondary mirror cluster may restart due to a stall in the log apply process.

    • Removed support for DES, 3DES and RC4 cipher suites.

    • Updated PostGIS extension to version 3.1.4.

    • Added support for postgis_raster extension version 3.1.4.

    PostgreSQL 10.17, Aurora PostgreSQL 2.9 (Deprecated on August 1, 2022. Make sure to upgrade to newer versions by January 31, 2023)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.17. For more information about the improvements in PostgreSQL 10.17, see PostgreSQL release 10.17.

    Aurora PostgreSQL 10.17.4, July 14, 2022

    Security enhancements

    • Backpatched the PostgreSQL community fix for CVE-2022-1552: Autovacuum, REINDEX, and others omit "security restricted operation". For more information, see CVE-2022-1552.

    High priority stability enhancements

    • Fixed an error handling issue related to out of memory conditions which could result in brief periods of unavailability.

    • Fixed an issue that could cause excess storage use due to files not being properly closed.

    • Fixed an issue that caused Performance Insights to show "Unknown wait event".

    Aurora PostgreSQL 2.9.2

    Security enhancements

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched postgis to PostGIS 2.4.7. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    Aurora PostgreSQL 2.9.1

    Critical stability updates

    • Fixed an issue where, in rare circumstances, a data cache of a read node may be inconsistent following a restart of that node.

    High priority stability updates

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    • Fixed an issue where Aurora may flag an issue following a major version update with the message: "PANIC: could not access status of next transaction id xxxxxxxx".

    Additional improvements and enhancements

    • Fixed an issue where read nodes restart due to a replication origin cache lookup failure.

    • Fixed an issue where in rare cases, an Aurora Global Database secondary mirror cluster may restart due to a stall in the log apply process.

    • Fixed an issue that causes Performance Insights to incorrectly set the backend type of a database connection.

    • Fixed an issue where orphaned files caused failed translations in read codepaths during or after major version upgrade.

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    • Fixed an out of memory crash issue with Aurora storage daemon that leads to writer node restart. This also reduces the overall system memory consumption.

    Aurora PostgreSQL 2.9

    High priority stability updates

    1. Fixed an issue where creating a database from an existing template database with tablespace resulted in an error with the message ERROR: could not open file pg_tblspc/...: No such file or directory.

    2. Fixed an issue where, in rare cases, an Aurora replica may be unable to start when a large number of PostgreSQL subtransactions (i.e. SQL savepoints) have been used.

    3. Fixed an issue where, in rare circumstances, read results may be inconsistent for repeated read requests on replica nodes.

    Additional improvements and enhancements

    1. Upgraded OpenSSL to 1.1.1k.

    2. Reduced CPU usage and memory consumption of the WAL apply process on Aurora replicas for some workloads.

    3. Improved safety checks in the write path to detect incorrect writes to metadata.

    4. Improved security by removing 3DES and other older ciphers for SSL/TLS connections.

    5. Fixed an issue where a duplicate file entry can prevent the Aurora PostgreSQL engine from starting up.

    6. Fixed an issue that could cause temporary unavailability under heavy workloads.

    7. Added back ability to use a leading forward slash in the S3 path during S3 import.

    8. Updated the orafce extension to version 3.16.

    9. Updated the PostGIS extension to version 2.4.7.

    PostgreSQL 10.16, Aurora PostgreSQL 2.8 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.16. For more information about the improvements in PostgreSQL 10.16, see PostgreSQL release 10.16.

    Aurora PostgreSQL 2.8.1

    Security enhancements

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched postgis to PostGIS 2.4.4. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched an input validation error in the log_fdw extension function parameters.

    Aurora PostgreSQL 2.8.0

    High priority stability updates

    1. Fixed a bug where in rare cases a reader had inconsistent results when it restarted while a transaction with more than 64 subtransactions was being processed.

    2. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    1. Fixed a bug where the database could not be started when there were many relations in memory-constrained environments.

    2. Fixed a bug in the apg_plan_mgmt extension that could cause brief periods of unavailability due to an internal buffer overflow.

    3. Fixed a bug on reader nodes that could cause brief periods of unavailability during WAL replay.

    4. Fixed a bug in the rds_activity_stream extension that caused an error during startup when attempting to log audit events.

    5. Fixed a bug that prevented minor version updates of an Aurora global database cluster.

    6. Fixed bugs in the aurora_replica_status function where rows were sometimes partially populated and some values such as Replay Latency, and CPU usage were always 0.

    7. Fixed a bug where the database engine would attempt to create shared memory segments larger than the instance total memory and fail repeatedly. For example, attempts to create 128 GiB shared buffers on a db.r5.large instance would fail. With this change, requests for total shared memory allocations larger than the instance memory allow setting the instance to incompatible parameters.

    8. Added logic to clean up unnecessary pg_wal temporary files on a database startup.

    9. Fixed a bug that reported ERROR: rds_activity_stream stack item 2 not found on top - cannot pop when attempting to create the rds_activity_stream extension.

    10. Fixed a bug that could cause the error failed to build any 3-way joins in a correlated IN subquery under an EXISTS subquery.

    11. Fixed a bug that could cause brief periods of unavailability due to running out of memory when creating the postgis extension with pgAudit enabled.

    12. Fixed a bug when using outbound logical replication to synchronize changes to another database that could fail with an error message like ERROR: could not map filenode "base/16395/228486645" to relation OID.

    13. Fixed a bug where the rds_ad role wasn't created after upgrading from a version of Aurora PostgreSQL that doesn't support Microsoft Active Directory authentication.

    14. Added btree page checks to detect tuple metadata inconsistency.

    15. Fixed a bug in asynchronous buffer reads that could cause brief periods of unavailability on reader nodes during WAL replay.

    PostgreSQL 10.14, Aurora PostgreSQL 2.7 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.14. For more information about the improvements in PostgreSQL 10.14, see PostgreSQL release 10.14.

    Aurora PostgreSQL 2.7.5

    Security enhancements

    • Modified the ip4r extension to mitigate a security issue during create extension. The issue was originally disclosed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched postgis to PostGIS 2.4.4. This is a PostGIS fix for the vulnerability addressed in core PostgreSQL by CVE-2020-14350. For more information, see CVE-2020-14350.

    • Backpatched an input validation error in the log_fdw extension function parameters.

    Aurora PostgreSQL 2.7.3

    High priority stability updates

    1. Provided a patch for PostgreSQL community security issues CVE-2021-32027, CVE-2021-32028 and CVE-2021-32029.

    Additional improvements and enhancements

    1. Fixed a bug in the aws_s3 extension to allow import of objects with leading forward slashes in the object identifier.

    2. Fixed a bug in the rds_activity_stream extension that caused an error during startup when attempting to log audit events.

    3. Fixed a bug that returned an ERROR when attempting to create the rds_activity_stream extension.

    4. Fixed a bug that could cause brief periods of unavailability due to running out of memory when creating the postgis extension with pgAudit enabled.

    5. Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    Aurora PostgreSQL 2.7.2

    High priority stability updates

    1. Fixed a bug where a reader node might render an extra or missing row if the reader restarted while the writer node is processing a long transaction with more than 64 subtransactions.

    Additional improvements and enhancements

    1. Fixed a bug that could lead to intermittent unavailability due to the rotation of network encryption keys.

    2. Fixed a bug where a large S3 import with thousands of clients can cause one or more of the import clients to stop responding.

    Aurora PostgreSQL 2.7.1

    Critical stability updates

    1. Fixed a bug that caused a read replica to unsuccessfully restart repeatedly in rare cases.

    2. Fixed a bug where a cluster became unavailable when attempting to create more than 16 read replicas or Aurora global database secondary Amazon Web Services Regions. The cluster became available again when the new read replica or secondary Amazon Web Services Region was removed.

    Additional improvements and enhancements

    1. Fixed a bug that when under heavy load, snapshot import, COPY import, or S3 import stopped responding in rare cases.

    2. Fixed a bug where a read replica might not join the cluster when the writer was very busy with a write-intensive workload.

    3. Fixed a bug that caused a cluster to take several minutes to restart if a logical replication stream was terminated while handling many complex transactions.

    4. Disallowed the use of both IAM and Kerberos authentication for the same user.

    Aurora PostgreSQL 2.7.0

    Critical stability updates

    • None

    High priority stability updates

    1. Backported fixes for the following PostgreSQL community security issues:

    2. Fixed a bug in Aurora PostgreSQL replication that could result in the error message ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    3. Fixed a bug where in some cases, DB clusters that have logical replication enabled did not remove truncated WAL segment files from storage. This resulted in volume size growth.

    4. Fixed a bug in the pg_stat_statements extension that caused excessive CPU consumption.

    Additional improvements and enhancements

    1. Improved the asynchronous mode performance of database activity streams.

    2. Aurora Serverless v1 for PostgreSQL now supports query execution on all connections during a scale event.

    3. Reduced the delay when publishing to CloudWatch the rpo_lag_in_msec metric for Aurora global database clusters.

    4. Fixed a bug in Serverless clusters where transaction processing was unnecessarily suspended for long periods when creating a scale point.

    5. Fixed a bug in Aurora Serverless v1 for PostgreSQL where a leaked lock resulted in a prolonged scale event.

    6. Fixed a bug in Aurora Serverless v1 for PostgreSQL where connections being migrated during a scale event was disconnected with the following message: ERROR: could not open relation with OID ...

    7. Aurora PostgreSQL no longer falls behind on a read node when the backend is blocked writing to the database client.

    8. Fixed a bug that in rare cases caused a brief period of unavailability on a read replica when the storage volume grew.

    9. Fixed a bug when creating a database that could return the following error: ERROR: could not create directory on local disk

    10. Fixed a bug where in some cases replaying XLOG_BTREE_REUSE_PAGE records on Aurora reader instances caused unnecessary replay lag.

    11. Fixed a bug in the GiST index that could result in an out of memory condition after promoting an Aurora read replica.

    12. Fixed a bug where the aurora_replica_status function showed truncated server identifiers.

    13. Fixed an S3 import bug that reported ERROR: HTTP 403. Permission denied when importing data from a file inside an S3 subfolder.

    14. Fixed a bug in the aws_s3 extension for pre-signed URL handling that could result in the error message S3 bucket names with a period (.) are not supported.

    15. Fixed a bug in the aws_s3 extension where an import might be blocked indefinitely if an exclusive lock was taken on the relation prior to beginning the operation.

    16. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance that uses GiST indexes. In rare cases, this bug caused a brief period of unavailability after promoting the Aurora cluster.

    17. Fixed a bug in database activity streams where customers were not notified of the end of an outage.

    PostgreSQL 10.13, Aurora PostgreSQL 2.6 (Deprecated on July 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.13. For more information about the improvements in PostgreSQL 10.13, see PostgreSQL release 10.13.

    Aurora PostgreSQL release 2.6.2

    Critical stability updates

    1. None

    High priority stability updates

    1. Fixed a bug in Aurora PostgreSQL replication that could result in the error message ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    2. Fixed a bug where in some cases, DB clusters that have logical replication enabled did not remove truncated WAL segment files from storage. This resulted in volume size growth.

    3. Fixed an issue with creating a global database cluster in a secondary region.

    4. Backported fixes for the following PostgreSQL community security issues:

    5. Fixed a bug in the pg_stat_statements extension that caused excessive CPU consumption.

    Additional improvements and enhancements

    1. Aurora PostgreSQL no longer falls behind on a read node when the backend is blocked writing to the database client.

    2. Reduced the delay when publishing to CloudWatch the rpo_lag_in_msec metric for Aurora global database clusters.

    3. Fixed a bug where a DROP DATABASE statement didn't remove any relation files.

    4. Fixed a bug where in some cases replaying XLOG_BTREE_REUSE_PAGE records on Aurora reader instances caused unnecessary replay lag.

    5. Fixed a small memory leak in a b-tree index that could lead to an out of memory condition.

    6. Fixed a bug in the aurora_replica_status() function where the server_id field was sometimes truncated.

    7. Fixed a bug where a log record was incorrectly processed causing the Aurora replica to crash.

    8. Fixed an S3 import bug that reported ERROR: HTTP 403. Permission denied when importing data from a file inside an S3 subfolder.

    9. Improved performance of the asynchronous mode for database activity streams.

    10. Fixed a bug in the aws_s3 extension that could result in the error message S3 bucket names with a period (.) are not supported.

    11. Fixed a race condition that caused valid imports to intermittently fail.

    12. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance that uses GiST indexes. In rare cases, this bug caused a brief period of unavailability after promoting the Aurora DB cluster.

    13. Fixed a bug in the aws_s3 extension where an import may be blocked indefinitely if an exclusive lock was taken on the relation prior to beginning the operation.

    Aurora PostgreSQL 2.6.1

    You can find the following improvements in this release.

    Critical stability enhancements

    1. Fixed a bug that appears when the NOT EXISTS operator incorrectly returns TRUE, which can only happen when the following unusual set of circumstances occurs:

      • A query is using the NOT EXISTS operator.

      • The column or columns being evaluated against the outer query in the NOT EXISTS subquery contain a NULL value.

      • There isn't a another predicate in the subquery that removes the need for the evaluation of the NULL values.

      • The filter used in the subquery does not use an index seek for its execution.

      • The operator isn't converted to a join by the query optimizer.

    Aurora PostgreSQL 2.6.0

    You can find the following improvements in this release.

    New features

    1. Added support for the RDKit extension version 3.8.

      The RDKit extension provides modeling functions for cheminformatics. Cheminformatics is storing, indexing, searching, retrieving, and applying information about chemical compounds. For example, with the RDKit extension you can construct models of molecules, search for molecular structures, and read or create molecules in various notations. You can also perform research on data loaded from the ChEMBL website or a SMILES file. The Simplified Molecular Input Line Entry System (SMILES) is a typographical notation for representing molecules and reactions. For more information, see The RDKit database cartridge in the RDKit documentation.

    2. Added support for the pglogical extension version 2.2.2.

      The pglogical extension is a logical streaming replication system that provides additional features beyond what's available in PostgreSQL native logical replication. Features include conflict handling, row filtering, DDL/sequence replication and delayed apply. You can use the pglogical extension to set up replication between Aurora PostgreSQL clusters, between RDS PostgreSQL and Aurora PostgreSQL, and with PostgreSQL databases running outside of RDS.

    3. Aurora dynamically resizes your cluster storage space. With dynamic resizing, the storage space for your Aurora DB cluster automatically decreases when you remove data from the DB cluster. For more information, see Storage scaling in the Amazon Aurora User Guide.

      Note

      The dynamic resizing feature is being deployed in phases to the Amazon Web Services Regions where Aurora is available. Depending on the Region where your cluster is, this feature might not be available yet. For more information, see the What's New announcement.

    Critical stability updates

    1. Fixed a bug related to heap page extend that in rare cases resulted in longer recovery time and impacted availability.

    High priority stability updates

    1. Fixed a bug when upgrading Aurora Global Database clusters from 10.11.

    2. Fixed a bug in Aurora Global Database that could cause delays in upgrading the database engine in a secondary Amazon Web Services Region. For more information, see Using Amazon Aurora global databases in the Amazon Aurora User Guide.

    3. Fixed a bug that in rare cases caused delays in upgrading a database to engine version 10.13.

    Additional improvements and enhancements

    1. Fixed a bug where the Aurora replica crashed when workloads with heavy subtransactions are made on the writer instance.

    2. Fixed a bug where the writer instance crashed due to a memory leak and the depletion of memory used to track active transactions.

    3. Fixed a bug that lead to a crash due to improper initialization when there is no free memory available during PostgreSQL backend startup.

    4. Fixed a bug where an Aurora PostgreSQL Serverless DB cluster might return the following error after a scaling event: ERROR: prepared statement "S_6" already exists.

    5. Fixed an out-of-memory problem when issuing the CREATE EXTENSION command with PostGIS when Database Activity Streams enabled.

    6. Fixed a bug where a SELECT query might incorrectly return the error Attempting to read past EOF of relation rrrr. blockno=bbb nblocks=nnn.

    7. Fixed a bug where the database might be unavailable briefly due to error handling in database storage growth.

    8. Fixed a bug in Aurora PostgreSQL Serverless where queries that executed on previously idle connections got delayed until the scale operation completed.

    9. Fixed a bug where an Aurora PostgreSQL DB cluster with Database Activity Streams enabled might report the beginning of a potential loss window for activity records, but does not report the restoration of connectivity.

    PostgreSQL 10.12, Aurora PostgreSQL 2.5 (Deprecated on March 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.12. For more information about the improvements in PostgreSQL 10.12, see PostgreSQL release 10.12.

    Aurora PostgreSQL 2.5.7

    You can find the following improvements in this release.

    Critical stability updates

    • None

    High priority stability updates

    1. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 2.5.6

    You can find the following improvements in this release.

    Critical stability updates

    • None

    High priority stability updates

    1. Fixed a bug in Aurora PostgreSQL replication that might result in the error message, ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    Additional improvements and enhancements

    1. Fixed a bug that in rare cases caused brief read replica unavailability when storage volume grew.

    2. Aurora PostgreSQL Serverless now supports execution of queries on all connections during a scale event.

    3. Fixed a bug in Aurora PostgreSQL Serverless where a leaked lock resulted in a prolonged scale event.

    4. Fixed a bug where the aurora_replica_status function showed truncated server identifiers.

    5. Fixed a bug in Aurora PostgreSQL Serverless where connections being migrated during a scale event disconnected with the message: "ERROR: could not open relation with OID ....

    6. Fixed a bug in a GiST index that might result in an out of memory condition after promoting an Aurora Read Replica.

    7. Improved performance for Database Activity Streams.

    8. Fixed a bug in Database Activity Streams where customers were not notified when an outage ended.

    9. Fixed a bug in the aws_s3 extension for pre-signed URL handling that could have resulted in the error message S3 bucket names with a period (.) are not supported.

    10. Fixed a bug in the aws_s3 extension where incorrect error handling could lead to failures during the import process.

    11. Fixed a bug in the aws_s3 extension where an import may be blocked indefinitely if an exclusive lock was taken on the relation prior to beginning the operation.

    Aurora PostgreSQL 2.5.4

    You can find the following improvements in this release.

    Critical stability enhancements

    1. Fixed a bug that appears when the NOT EXISTS operator incorrectly returns TRUE, which can only happen when the following unusual set of circumstances occurs:

      • A query is using the NOT EXISTS operator.

      • The column or columns being evaluated against the outer query in the NOT EXISTS subquery contain a NULL value.

      • There isn't a another predicate in the subquery that removes the need for the evaluation of the NULL values.

      • The filter used in the subquery does not use an index seek for its execution.

      • The operator isn't converted to a join by the query optimizer.

    Aurora PostgreSQL 2.5.3

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    • None

    Additional improvements and enhancements

    1. Fixed a bug in Aurora PostgreSQL Serverless where queries that ran on previously idle connections got delayed until the scale operation completed.

    2. Fixed a bug that might cause brief unavailability for heavy subtransaction workloads when multiple reader instances restart or rejoin the cluster.

    3. Fixed a bug in Aurora PostgreSQL Global Database where upgrading a secondary cluster might result in failure due to incorrect checksum versioning. This might have required re-creating the secondary clusters.

    Aurora PostgreSQL 2.5.2

    You can find the following improvements in this release.

    Critical stability updates

    1. Fixed a bug related to heap page extend that in rare cases resulted in longer recovery time and impacted availability.

    High priority stability updates

    1. Fixed a bug in Aurora Global Database that could cause delays in upgrading the database engine in a secondary region. For more information, see Using Amazon Aurora global databases in the Amazon Aurora User Guide.

    2. Fixed a bug that in rare cases caused delays in upgrading a database to engine version 10.12.

    Additional improvements and enhancements

    1. Fixed a bug where the database might be unavailable briefly due to error handling in database storage growth.

    2. Fixed a bug where a SELECT query might incorrectly return the error, Attempting to read past EOF of relation rrrr. blockno=bbb nblocks=nnn.

    3. Fixed a bug where an Aurora PostgreSQL Serverless DB cluster might return the following error after a scaling event: ERROR: prepared statement "S_6" already exists.

    Aurora PostgreSQL 2.5.1

    New features

    1. Added support for Amazon Aurora PostgreSQL Global Database. For more information, see Using Amazon Aurora global databases in the Amazon Aurora User Guide.

    2. Added the ability to configure the recovery point objective (RPO) of a global database for Aurora PostgreSQL. For more information, see Managing RPOs for Aurora PostgreSQL–based global databases in the Amazon Aurora User Guide.

    You can find the following improvements in this release.

    Critical stability updates

    None.

    High priority stability updates

    1. Improved performance and availability of read instances when applying DROP TABLE and TRUNCATE TABLE operations.

    2. Fixed a small but continuous memory leak in a diagnostic module that could lead to an out-of-memory condition on smaller DB instance types.

    3. Fixed a bug in the PostGIS extension which could lead to a database restart. This has been reported to the PostGIS community as https://trac.osgeo.org/postgis/ticket/4646.

    4. Fixed a bug where read requests might stop responding due to incorrect error handling in the storage engine.

    5. Fixed a bug that fails for some queries and results in the message ERROR: found xmin xxxxxx from before relfrozenxid yyyyyyy. This could occur following the promotion of a read instance to a write instance.

    6. Fixed a bug where an Aurora serverless DB cluster might crash while rolling back a scale attempt.

    Additional improvements and enhancements

    1. Improved performance for queries that read many rows from storage.

    2. Improved performance and availability of reader DB instances during heavy read workload.

    3. Enabled correlated IN and NOT IN subqueries to be transformed to joins when possible.

    4. Improved read performance of the pg_prewarm extension.

    5. Fixed a bug where an Aurora serverless DB cluster might report the message ERROR: incorrect binary data format in bind parameter ... following a scale event.

    6. Fixed a bug where a serverless DB cluster might report the message ERROR: insufficient data left in message following a scale event.

    7. Fixed a bug where an Aurora serverless DB cluster may experience prolonged or failed scale attempts.

    8. Fixed a bug that resulted in the message ERROR: could not create file "base/xxxxxx/yyyyyyy" as a previous version still exists on disk: Success. Please contact Amazon customer support. This can occur during object creation after PostgreSQL's 32-bit object identifier has wrapped around.

    9. Fixed a bug where the write-ahead-log (WAL) segment files for PostgreSQL logical replication were not deleted when changing the wal_level value from logical to replica.

    10. Fixed a bug in the pg_hint_plan extension where a multi-statement query could lead to a crash when enable_hint_table is enabled. This is tracked in the PostgreSQL community as https://github.com/ossc-db/pg_hint_plan/issues/25.

    11. Fixed a bug where JDBC clients might report the message java.io.IOException: Unexpected packet type: 75 following a scale event in an Aurora serverless DB cluster.

    12. Fixed a bug in PostgreSQL logical replication that resulted in the message ERROR: snapshot reference is not owned by resource owner TopTransaction.

    13. Changed the following extensions:

      • Updated orafce to version 3.8

    PostgreSQL 10.11, Aurora PostgreSQL 2.4 (Deprecated on March 15, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.11. For more information about the improvements in PostgreSQL 10.11, see PostgreSQL release 10.11.

    This release contains multiple critical stability updates. Amazon highly recommends upgrading your Aurora PostgreSQL clusters that use older PostgreSQL 10 engines to this release.

    Aurora PostgreSQL 2.4.4

    You can find the following improvements in this release.

    Critical stability updates

    • None

    High priority stability updates

    1. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 2.4.3

    New features

    1. Aurora PostgreSQL now supports the PostgreSQL vacuum_truncate storage parameter to manage vacuum truncation for specific tables. Set this storage parameter to false for a table to prevent the VACUUM SQL command from truncating the table's trailing empty pages.

    Critical stability updates

    • None

    High priority stability updates

    1. Fixed a bug where reads from storage might stop responding due to incorrect error handling.

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 2.4.2

    You can find the following improvements in this release.

    Critical stability updates

    1. Fixed a bug in which a reader DB instance might temporarily use stale data. This could lead to wrong results such as too few or too many rows. This error is not persisted on storage, and will clear when the database page containing the row has been evicted from cache. This can happen when the primary DB instance enters a transaction snapshot overflow due to having more than 64 subtransactions in a single transaction. Applications susceptible to this bug include those that use SQL savepoints or PostgreSQL exception handlers with more than 64 subtransactions in the top transaction.

    High priority stability updates

    1. Fixed a bug that may cause a reader DB instance to crash causing unavailability while attempting to the join the DB cluster. This can happen in some cases when the primary DB instance has a transaction snapshot overflow due to a high number of subtransactions. In this situation the reader DB instance will be unable to join until the snapshot overflow has cleared.

    Additional improvements and enhancements

    1. Fixed a bug that prevented Performance Insights from determining the query ID of a running statement.

    Aurora PostgreSQL 2.4.1

    You can find the following improvements in this release.

    Critical stability updates

    1. Fixed a bug in which the DB instance might be briefly unavailable due to the self-healing function of the underlying storage.

    High priority stability updates

    1. Fixed a bug that might cause the database engine to crash causing unavailability. This occurred if a newly established database connection encountered a resource exhaustion-related error during initialization after successful authentication.

    Additional improvements and enhancements

    1. Provided a fix for the pg_hint_plan extension that could lead the database engine to crash causing unavailability. The open source issue can be tracked at https://github.com/ossc-db/pg_hint_plan/pull/45.

    2. Fixed a bug where SQL of the form ALTER FUNCTION ... OWNER TO ... incorrectly reported ERROR: improper qualified name (too many dotted names).

    3. Improved the performance of GIN index vacuum via prefetching.

    4. Fixed a bug in open source PostgreSQL that could lead to a database engine crash causing unavailability. This occurred during parallel B-Tree index scans. This issue has been reported to the PostgreSQL community.

    5. Improved the performance of in-memory B-Tree index scans.

    6. Additional general improvements to the stability and availability of Aurora PostgreSQL.

    Aurora PostgreSQL 2.4.0

    You can find the following new features and improvements in this engine version.

    New features

    1. Support for exporting data to Amazon S3. For more information, see Exporting data from an Aurora PostgreSQL DB cluster to Amazon S3 in the Amazon Aurora User Guide.

    2. Support for Amazon Aurora Machine Learning. For more information, see Using machine learning (ML) with Aurora PostgreSQL in the Amazon Aurora User Guide.

    3. SQL processing enhancements include:

      • Optimizations for NOT IN with the apg_enable_not_in_transform parameter.

      • Semi-join filter pushdown enhancements for hash joins with the apg_enable_semijoin_push_down parameter.

      • Optimizations for redundant inner join removal with the apg_enable_remove_redundant_inner_joins parameter.

      • Improved ANSI compatibility options with the ansi_constraint_trigger_ordering, ansi_force_foreign_key_checks and ansi_qualified_update_set_target parameters.

      For more information, see Amazon Aurora PostgreSQL parameters in the Amazon Aurora User Guide.

    4. New and updated PostgreSQL extensions include:

    Critical stability updates

    1. Fixed a bug related to creating B-tree indexes on temporary tables that in rare cases may result in longer recovery time, and impact availability.

    2. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance. In rare cases, this bug causes a log write failure that may result in longer recovery time, and impact availability.

    3. Fixed a bug related to handling of reads with high I/O latency that in rare cases may result in longer recovery time, and impact availability.

    High priority stability updates

    1. Fixed a bug related to logical replication in which wal segments are not properly removed from storage. This can result in storage bloat. To monitor this, view the TransactionLogDiskUsage parameter.

    2. Fixed multiple bugs, which cause Aurora to crash during prefetch operations on Btree indexes.

    3. Fixed a bug in which an Aurora restart may timeout when logical replication is used.

    4. Enhanced the validation checks performed on data blocks in the buffer cache. This improves Aurora's detection of inconsistency.

    Additional improvements and enhancements

    1. The query plan management extension apg_plan_mgmt has an improved algorithm for managing plan generation for highly partitioned tables.

    2. Reduced startup time on instances with large caches via improvements in the buffer cache recovery algorithm.

    3. Improved the performance of the read-node-apply process under high transaction rate workloads by using changes to PostgreSQL LWLock prioritization. These changes prevent starvation of the read-node-apply process while the PostgreSQL ProcArray is under heavy contention.

    4. Improved handling of batch reads during vacuum, table scans, and index scans. This results in greater throughput and lower CPU consumption.

    5. Fixed a bug in which a read node may crash during the replay of a PostgreSQL SLRU-truncate operation.

    6. Fixed a bug where in rare cases, database writes may stall following an error returned by one of the six copies of an Aurora log record.

    7. Fixed a bug related to logical replication where an individual transaction larger than 1 GB in size may result in an engine crash.

    8. Fixed a memory leak on read nodes when cluster cache management is enabled.

    9. Fixed a bug in which importing an RDS PostgreSQL snapshot might stop responding if the source snapshot contains a large number of unlogged relations.

    10. Fixed a bug in which the Aurora storage daemon may crash under heavy I/O load.

    11. Fixed a bug related to hot_standby_feedback for read nodes in which the read node may report the wrong transaction id epoch to the write node. This can cause the write node to ignore the hot_standby_feedback and invalidate snapshots on the read node.

    12. Fixed a bug in which storage errors that occur during CREATE DATABASE statements are not properly handled. The bug left the resulting database inaccessible. The correct behavior is to fail the database creation and return the appropriate error to the user.

    13. Improved handling of PostgreSQL snapshot overflow when a read node attempts to connect to a write node. Prior to this change, if the write node was in a snapshot overflow state, the read node was unable to join. A message appeared in the PostgreSQL log file in the form DEBUG: recovery snapshot waiting for non-overflowed snapshot or until oldest active xid on standby is at least xxxxxxx (now yyyyyyy). A snapshot overflow occurs when an individual transaction has created over 64 subtransactions.

    14. Fixed a bug related to common table expressions in which an error is incorrectly raised when a NOT IN class exists in a CTE. The error is CTE with NOT IN fails with ERROR: could not find CTE CTE-Name.

    15. Fixed a bug related to an incorrect last_error_timestamp value in the aurora_replica_status table.

    16. Fixed a bug to avoid populating shared buffers with blocks belonging to temporary objects. These blocks correctly reside in PostgreSQL backend local buffers.

    17. Improved the performance of vacuum cleanup on GIN indexes.

    18. Fixed a bug where in rare cases Aurora may exhibit 100% CPU utilization while acting as a replica of an RDS PostgreSQL instance even when the replication stream is idle.

    19. Backported a change from PostgreSQL 11 which improves the cleanup of orphaned temporary tables. Without this change, it is possible that in rare cases orphaned temporary tables can lead to transaction ID wraparound. For more information, see this PostgreSQL community commit.

    20. Fixed a bug where a Writer instance may accept replication registration requests from Reader instances while having an uninitialized startup process.

    21. Changed the following extensions:

      • Updated pg_hint_plan to version 1.3.3.

      • Added plprofiler version 4.1.

    For information about extensions and modules, see Extensions supported for Aurora PostgreSQL 10.

    PostgreSQL 10.7, Aurora PostgreSQL 2.3 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 10.7 with the Aurora PostgreSQL 2.3 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.7. For more information about the improvements in PostgreSQL 10.7, see PostgreSQL release 10.7.

    Aurora PostgreSQL 2.3.5

    You can find the following improvements in this release.

    Improvements

    1. Fixed a bug that could cause DB instance restarts.

    2. Fixed a bug that could cause a crash when the PostgreSQL backend exits while using logical replication.

    3. Fixed a bug that could cause a restart when reads occurred during failovers.

    4. Fixed a bug with the wal2json module for logical replication.

    5. Fixed a bug that could result in inconsistent metadata.

    Aurora PostgreSQL 2.3.3

    You can find the following improvements in this release.

    Improvements

    1. Provided a backport fix for the PostgreSQL community security issue CVE-2019-10130.

    2. Provided a backport fix for the PostgreSQL community security issue CVE-2019-10164.

    3. Fixed a bug in which data activity streaming could consume excessive CPU time.

    4. Fixed a bug in which parallel threads scanning a B-tree index might stop responding following a disk read.

    5. Fixed a bug where use of the not in predicate against a common table expression (CTE) could return the following error: "ERROR: bad levelsup for CTE".

    6. Fixed a bug in which the read node replay process might stop responding while applying a modification to a generalized search tree (GiST) index.

    7. Fixed a bug in which visibility map pages could contain incorrect freeze bits following a failover to a read node.

    8. Optimized log traffic between the write node and read nodes during index maintenance.

    9. Fixed a bug in which queries on read nodes may crash while performing a B-tree index scan.

    10. Fixed a bug in which a query that has been optimized for redundant inner join removal could crash.

    11. The function aurora_stat_memctx_usage now reports the number of instances of a given context name.

    12. Fixed a bug in which the function aurora_stat_memctx_usage reported incorrect results.

    13. Fixed a bug in which the read node replay process could wait to stop conflicting queries beyond the configured max_standby_streaming_delay value.

    14. Additional information is now logged on read nodes when active connections conflict with the relay process.

    15. Provided a backport fix for the PostgreSQL community bug #15677, where a crash could occur while deleting from a partitioned table.

    Aurora PostgreSQL 2.3.1

    You can find the following improvements in this release.

    Improvements

    1. Fixed multiple bugs related to I/O prefetching that caused engine crashes.

    Aurora PostgreSQL 2.3.0

    You can find the following improvements in this release.

    New features

    1. Aurora PostgreSQL now performs I/O prefetching while scanning B-tree indexes. This results in significantly improved performance for B-tree scans over uncached data.

    Improvements

    1. Fixed a bug in which read nodes may fail with the error "too many LWLocks taken".

    2. Addressed numerous issues that caused read nodes to fail to startup while the cluster is under heavy write workload.

    3. Fixed a bug in which usage of the aurora_stat_memctx_usage() function could lead to a crash.

    4. Improved the cache replacement strategy used by table scans to minimize thrashing of the buffer cache.

    PostgreSQL 10.6, Aurora PostgreSQL 2.2 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 10.6 with the Aurora PostgreSQL 2.2 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.6. For more information about the improvements in PostgreSQL 10.6, see PostgreSQL release 10.6.

    Aurora PostgreSQL 2.2.1

    You can find the following improvements in this release.

    Improvements

    1. Improved stability of logical replication.

    2. Fixed a bug which could cause an error running queries. The message reported was of the form "CLOG segment 123 does not exist: No such file or directory".

    3. Increased the supported size of IAM passwords to 8KB.

    4. Improved consistency of performance under high throughput write workloads.

    5. Fixed a bug which could cause a read replica to crash during a restart.

    6. Fixed a bug which could cause an error running queries. The message reported was of the form "SQL ERROR: Attempting to read past EOF of relation".

    7. Fixed a bug which could cause an increase in memory usage after a restart.

    8. Fixed a bug which could cause a transaction with a large number of subtransactions to fail.

    9. Merged a patch from community PostgreSQL which addresses potential failures when using GIN indexes. For more information, see https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=f9e66f2fbbb49a493045c8d8086a9b15d95b8f18.

    10. Fixed a bug which could cause a snapshot import from RDS for PostgreSQL to fail.

    Aurora PostgreSQL 2.2.0

    You can find the following improvements in this release.

    New features

    1. Added the restricted password management feature. Restricted password management enables you to restrict who can manage user passwords and password expiration changes by using the parameter rds.restrict_password_commands and the role rds_password. For more information, see Restricting password management in the Amazon Aurora User Guide.

    PostgreSQL 10.5, Aurora PostgreSQL 2.1 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 10.5 with the Aurora PostgreSQL 2.1 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.5. For more information about the improvements in PostgreSQL 10.5, see PostgreSQL release 10.5.

    Aurora PostgreSQL 2.1.1

    You can find the following improvements in this release.

    Improvements

    1. Fixed a bug which could cause an error running queries. The message reported was of the form "CLOG segment 123 does not exist: No such file or directory".

    2. Increased the supported size of IAM passwords to 8KB.

    3. Improved consistency of performance under high throughput write workloads.

    4. Fixed a bug which could cause a read replica to crash during a restart.

    5. Fixed a bug which could cause an error running queries. The message reported was of the form "SQL ERROR: Attempting to read past EOF of relation".

    6. Fixed a bug which could cause an increase in memory usage after a restart.

    7. Fixed a bug which could cause a transaction with a large number of subtransactions to fail.

    8. Merged a patch from community PostgreSQL which addresses potential failures when using GIN indexes. For more information, see https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=f9e66f2fbbb49a493045c8d8086a9b15d95b8f18.

    9. Fixed a bug which could cause a snapshot import from RDS for PostgreSQL to fail.

    For information about extensions and modules, see Extensions supported for Aurora PostgreSQL 10.

    Aurora PostgreSQL 2.1.0

    You can find the following improvements in this release.

    New features

    1. General availability of Aurora Query Plan Management, which enables customers to track and manage any or all query plans used by their applications, to control query optimizer plan selection, and to ensure high and stable applicationperformance. For more information, see Managing query execution plans for Aurora PostgreSQL in the Amazon Aurora User Guide.

    2. Updated the libprotobuf extension to version 1.3.0. This is used by the PostGIS extension.

    3. Updated the pg_similarity extension to version 1.0.

    4. Updated the log_fdw extension to version 1.1.

    5. Updated the pg_hint_plan extension to version 1.3.1.

    Improvements

    1. Network traffic between the writer and reader nodes is now compressed to reduce network utilization. This reduces the chance of read node unavailability due to network saturation.

    2. Implemented a high performance, scalable subsystem for PostgreSQL subtransactions. This improves the performance of applications which make extensive use of savepoints and PL/pgSQL exception handlers.

    3. The rds_superuser role can now set the following parameters on a per-session, database, or role level:

      • log_duration

      • log_error_verbosity

      • log_executor_stats

      • log_lock_waits

      • log_min_duration_statement

      • log_min_error_statement

      • log_min_messages

      • log_parser_stats

      • log_planner_stats

      • log_replication_commands

      • log_statement_stats

      • log_temp_files

    4. Fixed a bug in which the SQL command "ALTER FUNCTION ... OWNER TO ..." might fail with error "improper qualified name (too many dotted names)".

    5. Fixed a bug in which a crash could occur while committing a transaction with more than two million subtransactions.

    6. Fixed a bug in community PostgreSQL code related to GIN indexes which can cause the Aurora Storage volume to become unavailable.

    7. Fixed a bug in which an Aurora PostgreSQL replica of an RDS for PostgreSQL instance might fail to start, reporting error: "PANIC: could not locate a valid checkpoint record".

    8. Fixed a bug in which passing an invalid parameter to the aurora_stat_backend_waits function could cause a crash.

    Known issues

    1. The pageinspect extension is not supported in Aurora PostgreSQL.

    PostgreSQL 10.4, Aurora PostgreSQL 2.0 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 10.4 with the Aurora PostgreSQL 2.0 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 10.4. For more information about the improvements in PostgreSQL 10.4, see PostgreSQL release 10.4.

    Aurora PostgreSQL 2.0.1

    You can find the following improvements in this release.

    Improvements

    1. Fixed a bug which could cause an error running queries. The message reported was of the form "CLOG segment 123 does not exist: No such file or directory".

    2. Increased the supported size of IAM passwords to 8KB.

    3. Improved consistency of performance under high throughput write workloads.

    4. Fixed a bug which could cause a read replica to crash during a restart.

    5. Fixed a bug which could cause an error running queries. The message reported was of the form "SQL ERROR: Attempting to read past EOF of relation".

    6. Fixed a bug which could cause an increase in memory usage after a restart.

    7. Fixed a bug which could cause a transaction with a large number of subtransactions to fail.

    8. Merged a patch from community PostgreSQL which addresses potential failures when using GIN indexes. For more information, see https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=f9e66f2fbbb49a493045c8d8086a9b15d95b8f18.

    9. Fixed a bug which could cause a snapshot import from RDS for PostgreSQL to fail.

    For information about extensions and modules, see Extensions supported for Aurora PostgreSQL 10.

    Aurora PostgreSQL 2.0.0

    You can find the following improvements in this release.

    Improvements

    1. This release contains all fixes, features, and improvements present in PostgreSQL 9.6.9, Aurora PostgreSQL 1.3 (Deprecated on October 28, 2021).

    2. The temporary file size limitation is user-configurable. You require the rds_superuser role to modify the temp_file_limit parameter.

    3. Updated the GDAL library, which is used by the PostGIS extension.

    4. Updated the ip4r extension to version 2.1.1.

    5. Updated the pg_repack extension to version 1.4.3.

    6. Updated the plv8 extension to version 2.1.2.

    7. Parallel queries – When you create a new Aurora PostgreSQL version 2.0 instance, parallel queries are enabled for the default.postgres10 parameter group. The parameter max_parallel_workers_per_gather is set to 2 by default, but you can modify it to support your specific workload requirements.

    8. Fixed a bug in which read nodes may crash following a specific type of free space change from the write node.

    PostgreSQL 9.6.22, Aurora PostgreSQL 1.11 (Deprecated on January 31, 2022)

    Note

    The PostgreSQL engine version 9.6.22 and Aurora PostgreSQL 1.10 are no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.22. For more information about the improvements in PostgreSQL 9.6.22, see PostgreSQL release 9.6.22.

    Aurora PostgreSQL 1.11.1

    High priority stability updates

    • Fixed an issue where queries may become unresponsive due to I/O resource exhaustion triggered by prefetch.

    Additional improvements and stability updates

    • Fixed multiple issues in the Aurora storage daemon that could lead to brief periods of unavailability when specific network configurations are used.

    Aurora PostgreSQL 1.11

    High priority stability enhancements

    1. Fixed an issue where creating a database from an existing template database with tablespace resulted in an error with the message ERROR: could not open file pg_tblspc/...: No such file or directory.

    2. Fixed an issue where, in rare cases, an Aurora replica may be unable to start when a large number of PostgreSQL subtransactions (i.e. SQL savepoints) have been used.

    3. Fixed an issue where, in rare circumstances, read results may be inconsistent for repeated read requests on replica nodes.

    Additional improvements and enhancements

    1. Upgraded OpenSSL to 1.1.1k.

    2. Reduced CPU usage and memory consumption of the WAL apply process on Aurora replicas for some workloads.

    3. Improve safety checks in the write path to detect incorrect writes to metadata.

    4. Fixed an issue where a duplicate file entry can prevent the Aurora PostgreSQL engine from starting up.

    5. Fixed an issue that could cause temporary unavailability under heavy workloads.

    6. Added back ability to use a leading forward slash in the S3 path during S3 import.

    7. Updated the PostGIS extension to version 2.4.7.

    8. Updated the Orafce extension to version 3.16.

    PostgreSQL 9.6.21, Aurora PostgreSQL 1.10 (Deprecated on January 31, 2022)

    Note

    The PostgreSQL engine version 9.6.21 and Aurora PostgreSQL 1.10 are no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.21. For more information about the improvements in PostgreSQL 9.6.21, see PostgreSQL release 9.6.21.

    Aurora PostgreSQL 1.10.0

    High priority stability enhancements

    1. Fixed a bug where in rare cases a reader had inconsistent results when it restarted while a transaction with more than 64 subtransactions was being processed.

    2. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    1. Fixed a bug where the database could not be started when there were many relations in memory-constrained environments.

    2. Fixed a bug in the apg_plan_mgmt extension that could cause brief periods of unavailability due to an internal buffer overflow.

    3. Fixed a bug where the database engine would attempt to create shared memory segments larger than the instance total memory and fail repeatedly. For example, attempts to create 128 GiB shared buffers on a db.r5.large instance would fail. With this change, requests for total shared memory allocations larger than the instance memory allow setting the instance to incompatible parameters.

    4. Added logic to clean up unnecessary pg_wal temporary files on a database startup.

    5. Fixed a bug in Aurora PostgreSQL 9.6 that sometimes prevented read/write nodes from starting up when inbound replication is used.

    6. Fixed a bug that could cause brief periods of unavailability due to running out of memory when creating the postgis extension with pgAudit enabled.

    7. Added btree page checks to detect tuple metadata inconsistency.

    PostgreSQL 9.6.19, Aurora PostgreSQL 1.9 (Deprecated on January 31, 2022)

    Note

    The PostgreSQL engine version 9.6.19 and Aurora PostgreSQL 1.9 are no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.19. For more information about the improvements in PostgreSQL 9.6.19, see PostgreSQL release 9.6.19.

    Aurora PostgreSQL 1.9.2

    High priority stability enhancements

    1. Fixed a bug where a reader node might render an extra or missing row if the reader restarted while the writer node is processing a long transaction with more than 64 subtransactions.

    Additional improvements and enhancements

    1. Fixed a bug where a large S3 import with thousands of clients can cause one or more of the import clients to stop responding.

    Aurora PostgreSQL 1.9.1

    Critical stability enhancements

    1. Fixed a bug that caused a read replica to unsuccessfully restart repeatedly in rare cases.

    Additional improvements and enhancements

    1. Fixed a bug that when under heavy load, snapshot import, COPY import, or S3 import stopped responding in rare cases.

    2. Fixed a bug where a read replica might not join the cluster when the writer was very busy with a write-intensive workload.

    Aurora PostgreSQL 1.9.0

    Critical stability enhancements

    • None

    High priority stability enhancements

    1. Backported a fix for the PostgreSQL community security issues CVE-2020-25694, CVE-2020-25695, and CVE-2020-25696.

    2. Fixed a bug in Aurora PostgreSQL replication that might result in the following error message: ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound

    Additional improvements and enhancements

    1. Aurora PostgreSQL no longer falls behind on a read node when the backend is blocked writing to the database client.

    2. Fixed a bug that in rare cases caused a brief period of unavailability on a read replica when the storage volume grew.

    3. Fixed a bug when creating a database that could return the following error: ERROR: could not create directory on local disk

    4. Fixed a bug in the GiST index that could result in an out of memory condition after promoting an Aurora read replica.

    5. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance that uses GiST indexes. In rare cases, this bug caused a brief period of unavailability after promoting the Aurora cluster.

    PostgreSQL 9.6.18, Aurora PostgreSQL 1.8 (Deprecated on January 31, 2022)

    Note

    The PostgreSQL engine version 9.6.18 and Aurora PostgreSQL 1.8 are no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.18. For more information about the improvements in PostgreSQL 9.6.18, see PostgreSQL release 9.6.18.

    There is no version 1.8.1.

    Aurora PostgreSQL release 1.8.2

    Critical stability enhancements

    1. None

    High priority stability enhancements

    1. Fixed a bug in Aurora PostgreSQL replication that could result in the error message ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    2. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    1. Aurora PostgreSQL no longer falls behind on a read node when the backend is blocked writing to the database client.

    2. Fixed a bug where a DROP DATABASE statement didn't remove any relation files.

    3. Fixed a small memory leak in a b-tree index that could lead to an out of memory condition.

    4. Fixed a bug in the aurora_replica_status() function where the server_id field was sometimes truncated.

    5. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance that uses GiST indexes. In rare cases, this bug caused a brief period of unavailability after promoting the Aurora DB cluster.

    Aurora PostgreSQL 1.8.0

    You can find the following improvements in this release.

    Critical stability enhancements

    1. Fixed a bug related to heap page extend that in rare cases resulted in longer recovery time and impacted availability.

    Additional improvements and enhancements

    1. Fixed a bug where the Aurora replica crashed when workloads with heavy subtransactions are made on the writer instance.

    2. Fixed a bug where the writer instance crashed due to a memory leak and the depletion of memory used to track active transactions.

    3. Fixed a bug that lead to a crash due to improper initialization when there is no free memory available during PostgreSQL backend startup.

    4. Fixed a crash during a BTree prefetch that occurred under certain conditions that depended on the shape and data contained in the index.

    5. Fixed a bug where a SELECT query might incorrectly return the error Attempting to read past EOF of relation rrrr. blockno=bbb nblocks=nnn.

    6. Fixed a bug where the database might be unavailable briefly due to error handling in database storage growth.

    PostgreSQL 9.6.17, Aurora PostgreSQL 1.7 (Deprecated on January 31, 2022)

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.17. For more information about the improvements in PostgreSQL 9.6.17, see PostgreSQL release 9.6.17.

    Aurora PostgreSQL 1.7.7

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    1. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 1.7.6

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    1. Fixed a bug in Aurora PostgreSQL replication that might result in the error message, ERROR: MultiXactId nnnn has not been created yet -- apparent wraparound.

    Additional improvements and enhancements

    1. Fixed a bug that in rare cases caused brief read replica unavailability when storage volume grew.

    2. Fixed a bug in a b-tree index read optimization that might have caused a brief period of unavailability.

    3. Fixed a bug in a GiST index that might result in an out of memory condition after promoting an Aurora Read Replica.

    Aurora PostgreSQL 1.7.3

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    • None

    Additional improvements and enhancements

    1. Fixed a bug that might cause brief unavailability for heavy subtransaction workloads when multiple reader instances restart or rejoin the cluster.

    Aurora PostgreSQL 1.7.2

    You can find the following improvements in this release.

    Critical stability enhancements

    1. Fixed a bug related to heap page extend that in rare cases resulted in longer recovery time and impacted availability.

    High Priority Stability Enhancements

    None

    Additional improvements and enhancements

    1. Fixed a bug where the database might be unavailable briefly due to error handling in database storage growth.

    2. Fixed a bug where a SELECT query might incorrectly return the error, Attempting to read past EOF of relation rrrr. blockno=bbb nblocks=nnn.

    3. Fixed an issue with the internal metrics collector that could result in erratic CPU spikes on database instances.

    Aurora PostgreSQL 1.7.1

    You can find the following improvements in this release.

    Critical stability enhancements

    None.

    High priority stability enhancements

    1. Improved performance and availability of read instances when applying DROP TABLE and TRUNCATE TABLE operations.

    2. Fixed a small but continuous memory leak in a diagnostic module that could lead to an out-of-memory condition on smaller DB instance types.

    3. Fixed a bug in the PostGIS extension which could lead to a database restart. This has been reported to the PostGIS community as https://trac.osgeo.org/postgis/ticket/4646.

    4. Fixed a bug where read requests might stop responding due to incorrect error handling in the storage engine.

    5. Fixed a bug that fails for some queries and results in the message ERROR: found xmin xxxxxx from before relfrozenxid yyyyyyy. This could occur following the promotion of a read instance to a write instance.

    Additional improvements and enhancements

    1. Improved performance for queries that read many rows from storage.

    2. Improved performance and availability of reader DB instances during heavy read workload.

    3. Fixed a bug that resulted in the message ERROR: could not create file "base/xxxxxx/yyyyyyy" as a previous version still exists on disk: Success. Please contact Amazon customer support. This can occur during object creation after PostgreSQL's 32-bit object identifier has wrapped around.

    4. Fixed a bug in the pg_hint_plan extension where a multi-statement query could lead to a crash when enable_hint_table is enabled. This is tracked in the PostgreSQL community as https://github.com/ossc-db/pg_hint_plan/issues/25.

    5. Changed the following extensions:

      • Updated orafce to version 3.8

    PostgreSQL 9.6.16, Aurora PostgreSQL 1.6 (Deprecated on January 31, 2022)

    This version of Aurora PostgreSQL is compatible with PostgreSQL 9.6.16. For more information about the improvements in release 9.6.16, see PostgreSQL release 9.6.16.

    This release contains multiple critical stability enhancements. Amazon highly recommends upgrading your Aurora PostgreSQL clusters that use older PostgreSQL 9.6 engines to this release.

    Aurora PostgreSQL 1.6.4

    You can find the following improvements in this release.

    Critical stability enhancements

    • None

    High priority stability enhancements

    1. Backported fixes for the following PostgreSQL community security issues:

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 1.6.3

    New features

    1. Aurora PostgreSQL now supports the PostgreSQL vacuum_truncate storage parameter to manage vacuum truncation for specific tables. Set this storage parameter to false when creating or altering a table to prevent the VACUUM SQL command from truncating the table's trailing empty pages.

    Critical stability enhancements

    • None

    High priority stability enhancements

    1. Fixed a bug where reads from storage might stop responding due to incorrect error handling.

    Additional improvements and enhancements

    • None

    Aurora PostgreSQL 1.6.2

    You can find the following improvements in this engine update.

    Critical stability enhancements

    1. Fixed a bug in which a reader DB instance might temporarily use stale data. This could lead to wrong results such as too few or too many rows. This error is not persisted on storage, and will clear when the database page containing the row has been evicted from cache. This can happen when the primary DB instance enters a transaction snapshot overflow due to having more than 64 subtransactions in a single transaction. Applications susceptible to this bug include those that use SQL savepoints or PostgreSQL exception handlers with more than 64 subtransactions in the top transaction.

    High priority stability enhancements

    1. Fixed a bug that may cause a reader DB instance to crash causing unavailability while attempting to the join the DB cluster. This can happen in some cases when the primary DB instance has a transaction snapshot overflow due to a high number of subtransactions. In this situation the reader DB instance will be unable to join until the snapshot overflow has cleared.

    Additional improvements and enhancements

    1. Fixed a bug that prevented Performance Insights from determining the query ID of a running statement.

    Aurora PostgreSQL 1.6.1

    You can find the following improvements in this engine update.

    Critical stability enhancements

    1. None

    High priority stability enhancements

    1. Fixed a bug that might cause the database engine to crash causing unavailability. This occurred if a newly established database connection encountered a resource exhaustion-related error during initialization after successful authentication.

    Additional improvements and enhancements

    1. Provided general improvements to the stability and availability of Aurora PostgreSQL.

    Aurora PostgreSQL 1.6.0

    You can find the following new features and improvements in this engine version.

    New features

    1. Updates to the apg_plan_mgmt extension. For more information, see Managing query execution plans for Aurora PostgreSQL in the Amazon Aurora User Guide.

    Critical stability enhancements

    1. Fixed a bug related to creating B-tree indexes on temporary tables that in rare cases may result in longer recovery time, and impact availability.

    2. Fixed a bug related to replication when Aurora PostgreSQL is acting as a physical replica of an RDS PostgreSQL instance. In rare cases, this bug causes a log write failure that may result in longer recovery time, and impact availability.

    3. Fixed a bug related to handling of reads with high I/O latency that in rare cases may result in longer recovery time, and impact availability.

    High priority stability enhancements

    1. Fixed multiple bugs, which cause Aurora to crash during prefetch operations on Btree indexes.

    2. Enhanced the validation checks performed on data blocks in the buffer cache. This improves Aurora's detection of inconsistency.

    Additional improvements and enhancements

    1. The query plan management extension apg_plan_mgmt has an improved algorithm for managing plan generation for highly partitioned tables.

    2. Reduced startup time on instances with large caches via improvements in the buffer cache recovery algorithm.

    3. Improved the performance of the read-node-apply process under high transaction rate workloads by using changes to PostgreSQL LWLock prioritization. These changes prevent starvation of the read-node-apply process while the PostgreSQL ProcArray is under heavy contention.

    4. Fixed a bug in which a read node may crash during the replay of a PostgreSQL SLRU-truncate operation.

    5. Fixed a bug where in rare cases, database writes might stall following an error returned by one of the six copies of an Aurora log record.

    6. Fixed a memory leak on read nodes when cluster cache management is enabled.

    7. Fixed a bug in which importing an RDS PostgreSQL snapshot might stop responding if the source snapshot contains a large number of unlogged relations.

    8. Fixed a bug related to hot_standby_feedback for read nodes in which the read node may report the wrong transaction id epoch to the write node. This can cause the write node to ignore the hot_standby_feedback and invalidate snapshots on the read node.

    9. Fixed a bug in which storage errors that occur during CREATE DATABASE statements are not properly handled. The bug left the resulting database inaccessible. The correct behavior is to fail the database creation and return the appropriate error to the user.

    10. Improved handling of PostgreSQL snapshot overflow when a read node attempts to connect to a write node. Prior to this change, if the write node was in a snapshot overflow state, the read node was unable to join. A message appear in the PostgreSQL log file in the form DEBUG: recovery snapshot waiting for non-overflowed snapshot or until oldest active xid on standby is at least xxxxxxx (now yyyyyyy). A snapshot overflow occurs when an individual transaction has created over 64 subtransactions.

    11. Fixed a bug related to common table expressions in which an error is incorrectly raised when a NOT IN class exists in a CTE. The error is CTE with NOT IN fails with ERROR: could not find CTE CTE-Name.

    12. Fixed a bug related to an incorrect last_error_timestamp value in the aurora_replica_status table.

    13. Fixed a bug to avoid populating shared buffers with blocks belonging to temporary objects. These blocks correctly reside in PostgreSQL backend local buffers.

    14. Fixed a bug where in rare cases Aurora may exhibit 100% CPU utilization while acting as a replica of an RDS PostgreSQL instance even when the replication stream is idle.

    15. Backported a change from PostgreSQL 11 which improves the cleanup of orphaned temporary tables. Without this change, it is possible that in rare cases orphaned temporary tables can to lead to transaction ID wraparound. For more information, see this PostgreSQL community commit.

    16. Fixed a bug where a Writer instance may accept replication registration requests from Reader instances while having an uninitialized startup process.

    17. Changed the following extensions:

      • Updated pg_hint_plan to version 1.2.5.

    PostgreSQL 9.6.12, Aurora PostgreSQL 1.5 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 9.6.12 with the Aurora PostgreSQL 1.5 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.12. For more information about the improvements in PostgreSQL 9.6.12, see PostgreSQL release 9.6.12.

    Aurora PostgreSQL 1.5.3

    You can find the following improvements in this release.

    Improvements

    1. Fixed a bug that could cause DB instance restarts.

    2. Fixed a bug that could cause a restart when reads occurred during failovers.

    3. Fixed a bug that could result in inconsistent metadata.

    Aurora PostgreSQL 1.5.2

    You can find the following improvements in this release.

    Improvements

    1. Provided a backport fix for the PostgreSQL community security issue CVE-2019-10130.

    2. Fixed a bug in which the read node replay process might stop responding while applying a modification to a generalized search tree (GiST) index.

    3. Fixed a bug in which visibility map pages may contain incorrect freeze bits following a failover to a read node.

    4. Fixed a bug in which the error "relation relation-name does not exist" is incorrectly reported.

    5. Optimized log traffic between the write node and read nodes during index maintenance.

    6. Fixed a bug in which queries on read nodes may crash while performing a B-tree index scan.

    7. The function aurora_stat_memctx_usage now reports the number of instances of a given context name.

    8. Fixed a bug in which the function aurora_stat_memctx_usage reported incorrect results.

    9. Fixed a bug in which the read node replay process may wait to stop conflicting queries beyond the configured max_standby_streaming_delay.

    10. Additional information is now logged on read nodes when active connections conflict with the relay process.

    Aurora PostgreSQL 1.5.1

    You can find the following improvements in this release.

    Improvements

    1. Fixed multiple bugs related to I/O prefetching, which caused engine crashes.

    Aurora PostgreSQL 1.5.0

    You can find the following improvements in this release.

    New features

    1. Aurora PostgreSQL now performs I/O prefetching while scanning B-tree indexes. This results in significantly improved performance for B-tree scans over uncached data.

    Improvements

    1. Addressed numerous issues that caused read nodes to fail to startup while the cluster is under heavy write workload.

    2. Fixed a bug in which usage of the aurora_stat_memctx_usage() function could lead to a crash.

    3. Improved the cache replacement strategy used by table scans to minimize thrashing of the buffer cache.

    PostgreSQL 9.6.11, Aurora PostgreSQL 1.4 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 9.6.11 with the Aurora PostgreSQL 1.4 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.11. For more information about the improvements in PostgreSQL 9.6.11, see PostgreSQL release 9.6.11.

    You can find the following improvements in this release.

    New features

    1. Support is added for the pg_similarity extension version 1.0.

    2. Aurora PostgreSQL now supports the PostgreSQL vacuum_truncate storage parameter to manage vacuum truncation for specific tables. Set this storage parameter to false when creating or altering a table to prevent the VACUUM SQL command from truncating the table's trailing empty pages.

    Improvements

    1. This release contains all fixes, features, and improvements present in PostgreSQL 9.6.9, Aurora PostgreSQL 1.3 (Deprecated on October 28, 2021).

    2. Network traffic between the writer and reader nodes is now compressed to reduce network utilization. This reduces the chance of read node unavailability due to network saturation.

    3. Performance of subtransactions has improved under high concurrency workloads.

    4. An update for the pg_hint_plan extension to version 1.2.3.

    5. Fixed an issue where on a busy system, a commit with millions of subtransactions (and sometimes with commit timestamps enabled) can cause Aurora to crash.

    6. Fixed an issue where an INSERT statement with VALUES could fail with the message "Attempting to read past EOF of relation".

    7. An upgrade of the apg_plan_mgmt extension to version 1.0.1. For details, see Version 1.0.1 of the Aurora PostgreSQL apg_plan_mgmt extension.

      The apg_plan_mgmt extension is used with query plan management. For more about how to install, upgrade, and use the apg_plan_mgmt extension, see Managing query execution plans for Aurora PostgreSQL in the Amazon Aurora User Guide.

    PostgreSQL 9.6.9, Aurora PostgreSQL 1.3 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 9.6.9 with the Aurora PostgreSQL 1.3 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    This release of Aurora PostgreSQL is compatible with PostgreSQL 9.6.9. For more information about the improvements in PostgreSQL 9.6.9, see PostgreSQL release 9.6.9.

    Aurora PostgreSQL 1.3.2

    You can find the following improvements in this release.

    New features

    1. Added the ProcArrayGroupUpdate wait event.

    Improvements

    1. Fixed a bug which could cause an error running queries. The message reported was of the form "CLOG segment 123 does not exist: No such file or directory".

    2. Increased the supported size of IAM passwords to 8KB.

    3. Improved consistency of performance under high throughput write workloads.

    4. Fixed a bug which could cause a read replica to crash during a restart.

    5. Fixed a bug which could cause an error running queries. The message reported was of the form "SQL ERROR: Attempting to read past EOF of relation".

    6. Fixed a bug which could cause an increase in memory usage after a restart.

    7. Fixed a bug which could cause a transaction with a large number of subtransactions to fail.

    8. Merged a patch from community PostgreSQL which addresses potential failures when using GIN indexes. For more information, see https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=f9e66f2fbbb49a493045c8d8086a9b15d95b8f18.

    9. Fixed a bug which could cause a snapshot import from RDS for PostgreSQL to fail.

    Aurora PostgreSQL 1.3.0

    You can find the following improvements in this release.

    Improvements

    1. This release contains all fixes, features, and improvements present in PostgreSQL 9.6.8, Aurora PostgreSQL 1.2 (Deprecated on October 28, 2021).

    2. Updated the GDAL library, which is used by the PostGIS extension.

    3. Updated the following PostgreSQL extensions:

      • ip4r updated to version 2.1.1.

      • pgaudit updated to version 1.1.1.

      • pg_repack updated to version 1.4.3.

      • plv8 updated to version 2.1.2.

    4. Fixed an issue in the monitoring system that could incorrectly cause a failover when local disk usage is high.

    5. Fixed a bug in which Aurora PostgreSQL can repeatedly crash, reporting:

      PANIC: new_record_total_len (8201) must be less than BLCKSZ (8192), rmid (6), info (32)

    6. Fixed a bug in which an Aurora PostgreSQL read node might be unable to rejoin a cluster due to recovery of a large buffer cache. This issue is unlikely to occur on instances other than r4.16xlarge.

    7. Fixed a bug in which inserting into an empty GIN index leaf page imported from pre-9.4 engine versions can cause the Aurora storage volume to become unavailable.

    8. Fixed a bug in which, in rare circumstances, a crash during transaction commit could result in the loss of CommitTs data for the committing transaction. The actual durability of the transaction was not impacted by this bug.

    9. Fixed a bug in the PostGIS extension in which PostGIS can crash in the function gserialized_gist_picksplit_2d().

    10. Improved the stability of read-only nodes during heavy write traffic on instances smaller than r4.8xl. This specifically addresses a situation where the network bandwidth between the writer and the reader is constrained.

    11. Fixed a bug in which an Aurora PostgreSQL instance acting as a replication target of an RDS for PostgreSQL instance crashed with the following error:

      FATAL: could not open file "base/16411/680897_vm": No such file or directory" during "xlog redo at 782/3122D540 for Storage/TRUNCATE"

    12. Fixed a memory leak on read-only nodes in which the heap size for the "aurora wal replay process" will continue to grow. This is observable via Enhanced Monitoring.

    13. Fixed a bug in which Aurora PostgreSQL can fail to start, with the following message reported in the PostgreSQL log:

      FATAL: Storage initialization failed.

    14. Fixed a performance limitation on heavy write workloads that caused waits on the LWLock:buffer_content and IO:ControlFileSyncUpdate events.

    15. Fixed a bug in which read nodes could crash following a specific type of free space change from the write node.

    PostgreSQL 9.6.8, Aurora PostgreSQL 1.2 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 9.6.8 with the Aurora PostgreSQL 1.2 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    For more information about PostgreSQL 9.6.8, see PostgreSQL release 9.6.8.

    Aurora PostgreSQL 1.2.2

    You can find the following improvements in this release.

    New features

    1. Added the ProcArrayGroupUpdate wait event.

    Improvements

    1. Fixed a bug which could cause an error running queries. The message reported was of the form "CLOG segment 123 does not exist: No such file or directory".

    2. Increased the supported size of IAM passwords to 8KB.

    3. Improved consistency of performance under high throughput write workloads.

    4. Fixed a bug which could cause a read replica to crash during a restart.

    5. Fixed a bug which could cause an error running queries. The message reported was of the form "SQL ERROR: Attempting to read past EOF of relation".

    6. Fixed a bug which could cause an increase in memory usage after a restart.

    7. Fixed a bug which could cause a transaction with a large number of subtransactions to fail.

    8. Merged a patch from community PostgreSQL which addresses potential failures when using GIN indexes. For more information, see https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=f9e66f2fbbb49a493045c8d8086a9b15d95b8f18.

    9. Fixed a bug which could cause a snapshot import from RDS for PostgreSQL to fail.

    Aurora PostgreSQL 1.2.0

    You can find the following improvements in this release.

    New features

    1. Introduced the aurora_stat_memctx_usage() function. This function reports internal memory context usage for each PostgreSQL backend. You can use this function to help determine why certain backends are consuming large amounts of memory.

    Improvements

    1. This release contains all fixes, features, and improvements present in PostgreSQL 9.6.6 Aurora PostgreSQL 1.1 (Deprecated on October 28, 2021).

    2. Updates the following PostgreSQL extensions:

      • pg_hint_plan updated to version 1.2.2

      • plv8 updated to version 2.1.0

    3. Improves efficiency of traffic between writer and reader nodes.

    4. Improves connection establishment performance.

    5. Improve the diagnostic data provided in the PostgreSQL error log when an out-of-memory error is encountered.

    6. Multiple fixes to improve the reliability and performance of snapshot import from Amazon RDS for PostgreSQL to Aurora PostgreSQL-Compatible Edition.

    7. Multiple fixes to improve the reliability and performance of Aurora PostgreSQL read nodes.

    8. Fixes a bug in which an otherwise idle instance can generate unnecessary read traffic on an Aurora storage volume.

    9. Fixes a bug in which duplicate sequence values can be encountered during insert. The problem only occurs when migrating a snapshot from RDS for PostgreSQL to Aurora PostgreSQL. The fix prevents the problem from being introduced when performing the migration. Instances migrated before this release can still encounter duplicate key errors.

    10. Fixes a bug in which an RDS for PostgreSQL instance migrated to Aurora PostgreSQL using replication can run out of memory doing insert/update of GIST indexes, or cause other issues with GIST indexes.

    11. Fixes a bug in which vacuum can fail to update the corresponding pg_database.datfrozenxid value for a database.

    12. Fixes a bug in which a crash while creating a new MultiXact (contended row level lock) can cause Aurora PostgreSQL to stop responding indefinitely on the first access to the same relation after the engine restarts.

    13. Fixes a bug in which a PostgreSQL backend can't be terminated or canceled while invoking an fdw call.

    14. Fixes a bug in which one vCPU is fully utilized at all times by the Aurora storage daemon. This issue is especially noticeable on smaller instance classes, such as r4.large, where it can lead to 25–50 percent CPU usage when idle.

    15. Fixes a bug in which an Aurora PostgreSQL writer node can fail over spuriously.

    16. Fixes a bug in which, in a rare scenario, an Aurora PostgreSQL read node can report:

      "FATAL: lock buffer_io is not held"

    17. Fixes a bug in which stale relcache entries can halt vacuuming of relations and push the system close to transaction ID wraparound. The fix is a port of a PostgreSQL community patch scheduled to be released in a future minor version.

    18. Fixes a bug in which a failure while extending a relation can cause Aurora to crash while scanning the partially extended relation.

    PostgreSQL 9.6.6 Aurora PostgreSQL 1.1 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 9.6.6 with the Aurora PostgreSQL 1.1 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    For more information about PostgreSQL 9.6.6 see, PostgreSQL release 9.6.6.

    You can find the following improvements in this engine update:

    New features

    1. Introduced the aurora_stat_utils extension. This extension includes two functions:

      • aurora_wait_report() function for wait event monitoring

      • aurora_log_report() for log record write monitoring

    2. Added support for the following extensions:

      • orafce 3.6.1

      • pgrouting 2.4.2

      • postgresql-hll 2.10.2

      • prefix 1.2.6

    Improvements

    1. This release contains all fixes, features, and improvements present in Aurora PostgreSQL 1.0.11

    2. Updates for the following PostgreSQL extensions:

      • postgis extension updated to version 2.3.4

      • geos library updated to version 3.6.2

      • pg_repack updated to version 1.4.2

    3. Access to the pg_statistic relation enabled.

    4. Disabled the 'effective_io_concurrency' guc parameter, as it does not apply to Aurora storage.

    5. Changed the 'hot_standby_feedback' guc parameter to not-modifiable and set the value to '1'.

    6. Improved heap page read performance during a vacuum operation.

    7. Improved performance of snapshot conflict resolution on read nodes.

    8. Improved performance of transaction snapshot acquisition on read nodes.

    9. Improved write performance for GIN meta page updates.

    10. Improved buffer cache recovery performance during startup.

    11. Fixes a bug that caused a database engine crash at startup while recovering prepared transactions.

    12. Fixes a bug that could result in the inability to start a read node when there are a large number of prepared transactions.

    13. Fixes a bug that could cause a read node to report:

      ERROR: could not access status of transaction 6080077

      DETAIL:* *Could not open file "pg_subtrans/005C": No such file or directory.

    14. Fixes a bug that could cause the error below when replicating from RDS PostgreSQL to Aurora PostgreSQL:

      FATAL: lock buffer_content is not held

      CONTEXT: xlog redo at 46E/F1330870 for Storage/TRUNCATE: base/13322/8058750 to 0 blocks flags 7

    15. Fixes a bug that could cause Aurora PostgreSQL to stop responding while replaying a multixact WAL record when replicating from RDS PostgreSQL to Aurora PostgreSQL.

    16. Multiple improvements to the reliability of importing snapshots from RDS PostgreSQL to Aurora PostgreSQL.

    PostgreSQL 9.6.3, Aurora PostgreSQL 1.0 (Deprecated on October 28, 2021)

    Note

    The PostgreSQL engine version 9.6.3 with the Aurora PostgreSQL 1.0 is no longer supported. To upgrade, see Upgrading the PostgreSQL DB engine for Aurora PostgreSQL in the Amazon Aurora User Guide.

    For more information about PostgreSQL 9.6.3 see, PostgreSQL release 9.6.3.

    This version includes the following Releases:

    Aurora PostgreSQL 1.0.11

    You can find the following improvements in this engine update:

    1. Fixes an issue with parallel query processing that can lead to incorrect results.

    2. Fixes an issue with visibility map handling during replication from Amazon RDS for PostgreSQL that can cause the Aurora storage volume to become unavailable.

    3. Corrects the pg-repack extension.

    4. Implements improvements to maintain fresh nodes.

    5. Fixes issues that can lead to an engine crash.

    Aurora PostgreSQL 1.0.10

    This update includes a new feature. You can now replicate an Amazon RDS PostgreSQL DB instance to Aurora PostgreSQL. For more information, see Replication with Amazon Aurora PostgreSQL in the Amazon Aurora User Guide.

    You can find the following improvements in this engine update:

    1. Adds error logging when a cache exists and a parameter change results in a mismatched buffer cache, storage format, or size.

    2. Fixes an issue that causes an engine reboot if there is an incompatible parameter value for huge pages.

    3. Improves handling of multiple truncate table statements during a replay of a write ahead log (WAL) on a read node.

    4. Reduces static memory overhead to reduce out-of-memory errors.

    5. Fixes an issue that can lead to out-of-memory errors while performing an insert with a GiST index.

    6. Improves snapshot import from RDS PostgreSQL, removing the requirement that a vacuum be performed on uninitialized pages.

    7. Fixes an issue that causes prepared transactions to return to the previous state following an engine crash.

    8. Implements improvements to prevent read nodes from becoming stale.

    9. Implements improvements to reduce downtime with an engine restart.

    10. Fixes issues that can cause an engine crash.

    Aurora PostgreSQL 1.0.9

    In this engine update, we fix an issue that can cause the Aurora storage volume to become unavailable when importing a snapshot from RDS PostgreSQL that contained uninitialized pages.

    Aurora PostgreSQL 1.0.8

    You can find the following improvements in this engine update:

    1. Fixes an issue that prevented the engine from starting if the shared_preload_libraries instance parameter contained pg_hint_plan.

    2. Fixes the error "Attempt to fetch heap block XXX is beyond end of heap (YYY blocks)," which can occur during parallel scans.

    3. Improves the effectiveness of prefetching on reads for a vacuum.

    4. Fixes issues with snapshot import from RDS PostgreSQL, which can fail if there are incompatible pg_internal.init files in the source snapshot.

    5. Fixes an issue that can cause a read node to crash with the message "aurora wal replay process (PID XXX) was terminated by signal 11: Segmentation fault". This issue occurs when the reader applied a visibility map change for an uncached visibility map page.

    Aurora PostgreSQL 1.0.7

    This is the first generally available release of Amazon Aurora PostgreSQL-Compatible Edition.