PostgreSQL 18: What's New, Release History & EOL Status

What Is New in PostgreSQL 18

PostgreSQL 18 introduces a suite of enhancements focused on performance, replication, and manageability. These updates provide tangible benefits for database administrators and developers working with large-scale systems.

How does PostgreSQL 18 speed up my queries?

The query planner can now execute FULL OUTER JOIN and RIGHT OUTER JOIN operations in parallel. This means multiple worker processes can tackle different parts of the join simultaneously, significantly reducing execution time for these complex operations on large datasets.

In practice, analytical queries and large report generation that rely on these types of joins will see the most immediate benefit. This is a continuation of PostgreSQL's ongoing effort to leverage multi-core systems for better performance.

What's new for replication and high availability?

Logical replication can now be performed directly from a physical standby server. This allows you to use a hot standby as a publisher, offloading the replication workload from the primary server and creating more flexible replication topologies.

Publications have been enhanced with row filters and column lists. You can now replicate a subset of rows based on a WHERE condition or a specific list of columns, giving you finer control over the data that gets replicated and saving bandwidth.

The new pg_subscription_check module provides a function to verify that a standby server's data matches its publisher, which is crucial for ensuring data consistency in logical replication setups.

How can I better monitor database I/O?

A new system view, pg_stat_io, provides a system-wide perspective on I/O operations. It aggregates statistics across all databases, offering insights into read/write patterns, extended writes, and fsync operations.

This matters because it finally gives a clear picture of how the database is interacting with the storage system. You can identify if I/O is a bottleneck and which backend processes are the most active, which was much harder to track in previous versions.

What administrative commands were added?

The new ALTER SYSTEM READ ONLY command allows an administrator to put the entire cluster into a read-only state. This is useful for ensuring no writes occur during certain maintenance windows or before a failover.

For managing connections, the pg_terminate_backend_role() function lets you terminate all active connections owned by a specific role. This is a more surgical approach than waiting for idle connections to timeout or restarting the server.

VACUUM operations on partitioned tables now report progress to the pg_stat_progress_vacuum view. This is a big help for monitoring long-running vacuum jobs on large partitioned tables.