A live PostgreSQL 19 terminal in the Certuvia sandbox demonstrates REPACK and autovacuum score queries.
*PostgreSQL 19 lands on 19 November 2026 with a suite of developer‑focused SQL additions and low‑level storage tweaks. The Certuvia sandbox lets operators test REPACK, autovacuum priority scores and logical replication on real servers.*
PostgreSQL 19 arrived on 19 November 2026, promising less friction and more control for database operators. Certuvia’s interactive blog post turns the release announcement into a live lab, deploying a real PostgreSQL 19 server in a disposable guest. Users can test new SQL constructs, storage‑level rewrites, and replication setups without administrative credentials. The sandbox runs under the student role, exposing a prepared demo database called blog_lab. Each experiment is isolated, with logical and physical replicas launched on demand. The environment mirrors production constraints: only one replica runs at a time, and memory limits force sequential testing. This hands‑on approach lets engineers verify claims before committing to an upgrade.
The new release adds IGNORE NULLS, get‑or‑create inserts, JSON export, JSONPath transformations and range subtraction. Each feature links to official documentation and can be tried in the sandbox. The release notes list the full change log, but the hands‑on tour highlights these as the most immediately useful for developers. The additions sit alongside planner advice improvements, giving query optimizers more levers. All examples run on a real PostgreSQL 19 server, not a simulated environment.
PostgreSQL 19 introduces a native REPACK command that rewrites tables and returns unused bytes to the operating system. The demo deletes rows, runs REPACK (CONCURRENTLY, ANALYZE) on the events table, and compares file size before and after. Unlike the older VACUUM (FULL, ANALYZE) which blocks, REPACK can run concurrently, briefly locking only at the final swap. PostgreSQL 18 lacks a native concurrent variant; users must install the pg_repack extension. The sandbox shows that REPACK requires the table's MAINTAIN privilege, which the student role already has.
Version 19 adds real priority scores to autovacuum, visible via pg_stat_autovacuum_scores. Scores reflect workload and statistics, allowing the scheduler to focus on the most critical tables. The sandbox’s autovacuum_max_parallel_workers is set to zero, so the demo inspects scores without launching parallel workers. Autovacuum can also mark eligible pages all‑visible during scans, improving future index‑only scans. PostgreSQL 18 does not expose these scores, and its scheduler cannot be switched back to the older behavior.
The Certuvia environment spins up a logical subscriber on port 5433 and a physical standby on port 5434, both fed by the primary on 5432. Only one replica runs at a time to fit the guest’s memory budget; starting a new card stops the previous instance. The logical replication walkthrough shows server setup and SQL directly, using the Linux postgres account for utilities. The sandbox lets users reset all three clusters, re‑initialize them, and rerun experiments in any order.
PostgreSQL 19 adds IGNORE NULLS, get‑or‑create inserts, JSON export, JSONPath transformations, and range subtraction, all documented in the official release notes.
REPACK rewrites tables and returns unused bytes to the OS, can run concurrently, and locks only briefly at swap, whereas VACUUM (FULL) blocks and does not reclaim space to the OS.
Yes, version 19 introduces real priority scores visible via pg_stat_autovacuum_scores, allowing the scheduler to focus on high‑need tables, a feature absent in PostgreSQL 18.
The Certuvia sandbox strips away marketing gloss, delivering raw evidence of PostgreSQL 19’s claims. REPACK, autovacuum scores, and the expanded SQL toolkit are now testable on a live server. For teams weighing an upgrade, the sandbox offers a clear, reproducible path to validate performance and operational impact before any production rollout.
Sources: Hacker News, Certuvia Blog