PostgreSQL 15

PostgreSQL 15 added SQL-standard MERGE, allowing conditional insert, update, or delete work to be expressed in one statement. It expanded logical replication with row filters and column lists in publications, making it possible to replicate selected rows or columns without separate replication middleware. It added more compression choices in several areas, including LZ4 and Zstandard support for WAL compression, and improved backup compression options. Query execution and maintenance also gained performance improvements, including faster sorting in common cases, better parallel execution for some queries, and improvements to vacuum and index-related work. New clusters created by this line use more restrictive default permissions for the public schema than earlier defaults; upgraded clusters retain their existing permissions unless administrators change them.

PostgreSQL 15 remains a supported and maintainable production choice when it is kept current on its minor-release stream. Staying on it does not prevent receipt of fixes that are backported to that stream, but it does leave later major-version capabilities, performance work, and operational improvements unavailable. Teams can also accumulate upgrade risk if application drivers, extensions, operating-system packages, or managed-service offerings move their primary testing toward newer PostgreSQL lines. The main operational risk is not PostgreSQL 15 itself, but treating a supported major version as a reason to defer compatibility testing until the eventual migration is larger and more constrained.

Keep every PostgreSQL 15 instance on the current minor release offered for this major line, and apply updates through a tested maintenance process. Inventory installed extensions, custom C-language functions, drivers, backup tooling, replication consumers, and monitoring agents, then verify their compatibility with the intended newer PostgreSQL target. Rehearse an upgrade on a production-sized copy using pg_upgrade where downtime requirements permit; otherwise evaluate logical replication or a managed-service migration path. Validate application behavior around MERGE, logical replication filters, collation-dependent indexes, backup and restore, and public-schema privileges, especially when creating new databases or restoring into newly initialized clusters. Record measured upgrade duration, rollback steps, and post-upgrade checks so the move can be scheduled before dependency constraints force it.