Added the cfs_compression configuration parameter that allows you to temporary disable CFS compression. It speeds up the migration of large compressed tables.

Fixed an issue with CFS failing to compress relation files where OID exceeds INT32_MAX. The issue occurred when moving an existing table to a CFS tablespace.

Improved the mechanism of aqo learning on subqueries.

Fixed an issue where the aqo_cleanup function did not remove data in the basic mode.

Fixed an issue where the leader was lost after non-incremental upgrade of a BiHA cluster from Postgres Pro Enterprise 17.4 or lower to 17.6.

Fixed a bug that occurred when the MIN value of the synchronous_standby_names configuration parameter was higher than 0 and the current number of standbys was higher than the num_sync value. This could cause hanging of transactions in case of replication hanging.

Fixed a bug where biha.add_to_ssn, biha.remove_from_ssn, and biha.set_ssn functions modified the list of synchronous standbys of the synchronous_standby_names parameter on the leader node only.

Fixed a bug where bihactl init --convert failed when executed in a Postgres Pro Enterprise instance with configured pg_proaudit.

Fixed an issue of multiple error messages caused by large heartbeats that might not fit into the shared memory buffer. The issue was solved by decreasing the size of heartbeats written to WAL.

Fixed memory leaks caused by incorrect error handling.

Fixed redundant memory allocation checks.

Fixed an issue where multimaster returned an error when executing read-write transactions containing queries to postgres_fdw foreign tables. The root cause was that postgres_fdw does not support two-phase commit transactions used by multimaster. To fix the issue, the postgres_fdw.read_only_transactions configuration parameter was implemented. When set to on, it enables read-only queries to foreign tables in read-write transactions, which allows seamless operation of multimaster and postgres_fdw.

Fixed an issue of potential transaction commit latency caused by an error in the exchange logic for messages containing information about transaction application on nodes.

Implemented the ability to re-read hint strings if hints were added outside the standard planner. This functionality is required to provide interoperability with the pgpro_multiplan extension when working with plans that are based on hint sets.

Fixed an issue in pg_hint_plan that could result in an infinite loop while processing queries with dollar-quoted strings.

Added support for logging of the following event types: ANALYZE, DISCARD ALL, DISCARD PLANS, DISCARD SEQUENCES, DISCARD TEMP (for temporary tables), CHECKPOINT, FETCH, and MOVE.

Added the new event class MISC. This class allows logging of all events (including those mentioned above) not related to other event classes.

Added a new method for logging security events. Events in CEF format can now be logged not only in a separate file but also in syslog.

Added a new event log field: application name. AUTHENTICATE events now contain the name of the initiating application alongside the connection parameters.

Fixed an issue with pg_proaudit failing to log security events in a CEF file.

Fixed an issue with pg_proaudit failing to log events for the INSERT command following CREATE within anonymous code blocks (the DO command) and stored procedures/functions.

Fixed an issue with pg_proaudit overloading a CPU core due to failed log rotation.

Added support for Postgres Pro Enterprise 18.

Added support for Transparent Data Encryption (TDE). The version output of pg_probackup now displays a TDE support indicator.

Made the host from which a backup was taken to be saved in backup.control and displayed in the show command output.

Improved the behavior during restore from read-only media, including more informative warning messages.

For archive-push and archive-get operations in the remote mode, made SIGQUIT signals to be ignored to prevent core dumps of helper programs when the server stops.

Added validation to ensure consistency between the port number in PGDATA and the --pgport option. Also added the --force flag to perform backups even when these values mismatch.

Fixed handling of external tablespaces during incremental restore. When using the --force flag, the tablespace contents are no longer overwritten.

Fixed a problem with the backup command holding the replication slot after successful completion.

Fixed the identifier conflict for the --wal-tree option by adding a uniqueness test.

Fixed the order of displaying the calculated and expected checksum values in the warning message.

Implemented the ability to store global prepared statements and use them in multiple sessions within a single server.

Improved the stability of the auto_approve_plans mode for real-time query replanning with the enabled pgpro_multiplan.auto_capturing parameter.

Added the new pgpro_multiplan_remove_template function that removes the specified template plan from the permanent pgpro_multiplan storage.

Added the new pgpro_multiplan_template_hintset_update function that replaces the generated hint set for the template plan with the specified set of custom hints.

Improved messages for errors that can occur when registering queries and freezing query plans.

Changed the logic of specifying values of the pgpro_multiplan.wildcards parameter. Now this parameter accepts a single regular expression rather than a comma-separated list of multiple expressions.

Fixed an issue where the template plan was not identified and applied if constants in the corresponding query were changed.

Fixed an issue in pgpro_multiplan that could cause a server crash when the same query was executed repeatedly due to incorrect handling of the sample_string value.

Improved take_sample time tracking.

Added support of pgpro_stats 1.10.

Added a possibility to exclude server connection strings from dumps.

Added statistics on restartpoints to the reports when it is available.

Added new functions for exchange management and procedures for message publishing to support efficient message distribution based on the publish-subscribe model.

Added new procedures for queue access management: reset_queue_access, grant_queue_access, and revoke_queue_access. Queue management procedures can now be used only by a superuser who created the pgpro_queue extension, while the grant_queue_access and revoke_queue_access procedures are also available to the queue owner.

Implemented automatic cache invalidation at the cluster level using WAL files. It can be enabled using the pgpro_result_cache.consistent and pgpro_result_cache.wal parameters.

Added a possibility to set the pgpro_result_cache.max_memory_size parameter at runtime using the ALTER SYSTEM command.

Added support for the new no_result_cache hint that allows excluding specific queries from caching.

Added two optional limit and offset arguments to the result_cache hint. These arguments allow you to restrict the number of rows returned from cached result sets and skip certain rows before returning.

Fixed an issue that could occur in pgpro_result_cache when caching a large number of simple queries by changing the algorithm for storing cached results.

Fixed a segmentation fault that could occur when reloading the configuration file after the pgpro_result_cache.consistent parameter was set to on.

Added new statistics to pgpro_stats_statements and pgpro_stats_totals views. These statistics were added to pg_stats_statements in Postgres Pro 18. Related functions are updated accordingly.

Added the new pgpro_stats_statements_filtered function to filter statistics similar to those provided in the pgpro_stats_statements view by one or several fields: userid, dbid, queryid, and planid. Calling this function is faster than querying the pgpro_stats_statements view with appropriate values of these fields.

Improved the behavior of pgpro_stats similarly to pg_stat_statements. The changes are:

Fixed an issue that could occur when in pgpro_stats custom metrics were changed on the fly. Now any change to custom metrics requires a server restart.

Provided interoperability with the Compressed File System (CFS) feature. pgpro_tde can now encrypt compressed data blocks. Note, however, that pg_rewind cannot be currently used for tablespaces that are both encrypted and compressed.

Optimized WAL records processing to improve performance.

Implemented the weighted-round-robin algorithm of workload distribution, as well as the proxima.load_balancer_node_weight configuration parameter to set node weights.

Implemented the least-connections algorithm of workload distribution.

Implemented integration of proxima and pgpro_multiplan to avoid establishing the dynamic dedicated session for queries with prepared statements. For more information, refer to Integration with pgpro_multiplan for Prepared Statements.

Added the pg_backend_set_config function to the list of objects and functions that make proxima automatically establish the dynamic dedicated session. To view the full list, refer to Dynamic Dedicated Session.

Implemented code optimization and fixed potential issues of the dynamic dedicated session functionality.

Implemented a range of internal fixes and optimizations to improve performance and reliability of the proxima extension.

Fixed a bug where the dynamic dedicated session established by the SET TRANSACTION command could not be terminated.

Fixed an issue where proxima would hang causing high CPU usage and disk overflow. The hanging occurred when attempting to connect to a follower in a BiHA cluster where nodes had different port numbers set for the proxima.port configuration parameter. Currently, the hanging was fixed, however, the proxima.port value must be identical on all nodes for normal operation of the BiHA cluster.