PostgreSQL 14

PostgreSQL 14 added multirange types, allowing a value to represent multiple non-overlapping ranges, and added SQL-standard SEARCH and CYCLE clauses for recursive common table expressions. It introduced LZ4 compression for oversized TOAST-stored values when PostgreSQL is built with LZ4 support. Query execution gained Memoize, which can cache results from parameterized scans in suitable nested-loop plans, and VACUUM can clean multiple eligible indexes in parallel. Logical replication can stream large transactions before they commit, and libpq pipeline mode lets client applications send multiple commands without waiting for each individual result.

PostgreSQL 14 remains within its supported window, so staying on it does not by itself mean running without upstream fixes; teams should still apply current minor releases. The trade-off is that later PostgreSQL lines contain further planner, replication, maintenance, observability, and operational improvements that are unavailable on 14. Over time, newer client libraries, extensions, managed-service options, and internal platform standards may focus testing on later server versions, increasing compatibility testing and upgrade urgency. Deferring the work also leaves less time to find query-plan changes, extension issues, collation differences, and application assumptions before an eventual major-version migration.

Keep every PostgreSQL 14 instance on the latest available minor release and verify that backups can be restored, including required roles, extensions, and external archive tooling. Inventory extensions, drivers, replication topology, collations, and any use of deprecated application behavior, then confirm their compatibility with the intended newer PostgreSQL major version. Build a representative test environment and rehearse the chosen upgrade method: pg_upgrade for a short controlled outage, or logical replication for a lower-downtime cutover where its limitations fit the workload. Compare application results and performance before and after the test upgrade, paying particular attention to execution plans, long-running transactions, replication lag, extension upgrades, and collation-dependent indexes. Use the rehearsal to define rollback steps, downtime or cutover criteria, and a production migration window.