YuSMP Group logo
YuSMP Group Editorial Team Editorial team, YuSMP Group
Dark data center aisle with one open server rack in the center, its drive trays lit in cyan and rearranged into a neat order by glowing light lines, with a laptop on a small cart showing colored bars

The short version

PostgreSQL 19 is a solid upgrade target, but treat it as a planned migration, not a drop-in patch. The biggest wins for application teams running PostgreSQL in production are online table repacking and read-your-writes on replicas. The biggest risks are changed defaults and features that were pulled late. Test on a copy of production first.

Why did PostgreSQL 19 ship late?

PostgreSQL 19 shipped late because AI-assisted code review found more defects than the project could fix before its usual autumn date. The release was first expected in late September or early October. In October the project moved general availability to November 19.

Peter Eisentraut, a PostgreSQL committer and vice president at EDB, told The Register that LLM-assisted review finds problems much faster, and teams end up “staring at a list of forty defect reports.” He said the cause was not careless code. Better tooling raised the quality bar in the middle of the cycle.

The same pressure shows in the November 12 update releases. They fixed 52 security vulnerabilities and more than 140 bugs across versions 14 to 18, a record for a quarterly release, with AI-assisted scanning behind many of the reports.

What is new in PostgreSQL 19?

PostgreSQL 19 focuses on running busy databases without downtime and on small developer conveniences. The changes most teams will notice:

AreaWhat changedWhy it matters
Table maintenanceREPACK (CONCURRENTLY) reclaims space and can reorder a table to match an indexRemoves bloat without blocking reads or writes, which used to need a maintenance window or the pg_repack extension
Query planspg_plan_advice and pg_stash_advice extensionsPin a known-good plan for a query and apply it automatically by query ID
ReplicasWAIT FOR LSN commandRead-your-writes on a standby, so read traffic can move off the primary more safely
ReplicationSequences replicate with FOR ALL SEQUENCES; logical WAL level turns on with the first slotFewer manual steps in blue-green upgrades and migrations; no restart to enable logical replication
SQLINSERT ... ON CONFLICT DO SELECT; IGNORE NULLS for lead, lag and related window functions“Get or create” in one round trip
Data exportCOPY can write JSON, including a single JSON arraySimpler exports to APIs and data pipelines
PerformanceForeign key checks make bulk inserts nearly 2x faster; parallel index vacuuming in autovacuumFaster loads and less vacuum lag on large tables
SecurityServer-side SNI with a new pg_hosts.confOne server can present different TLS certificates per hostname

What did not make it into the release?

The project reverted 8 to 10 significant features late in the cycle, The Register reported, and SQL Property Graph Queries (SQL/PGQ) is the best known of them. If a roadmap item depended on native graph queries in PostgreSQL 19, move it to a later major version or keep using an extension or a separate graph store.

Which defaults could break an upgrade?

Several defaults changed, and a straight upgrade can alter behavior without any error:

  • JIT is off by default. The project cites unreliable cost estimates. Analytical workloads that relied on JIT may run slower until you turn it back on for them.
  • RADIUS authentication is gone. Clusters that use it must move to another method before upgrading.
  • MD5 passwords warn on login. MD5 has been deprecated since PostgreSQL 18 and will be removed later. Switch roles to SCRAM now.
  • I/O worker settings changed. io_workers is replaced by io_min_workers and io_max_workers, so check configuration files and infrastructure-as-code templates.
  • TOAST compression defaults to lz4 when the server is built with LZ4 support.

What it means for US & EU software teams

For teams on PostgreSQL 14, the question is no longer whether to upgrade. Version 14 stopped receiving fixes with 14.25, the same month the project patched a record 52 vulnerabilities. That matters for audits: SOC 2, PCI DSS and EU frameworks such as NIS2 and DORA expect supported, patched software in production.

For teams on 15 to 18, PostgreSQL 19 is worth planning rather than rushing. REPACK and WAIT FOR LSN solve real operational problems, especially for SaaS and fintech products with large, write-heavy tables. A first major release still carries more risk than its later minor versions, and this one shipped after an unusual number of late fixes.

Managed services add their own lag. Cloud providers usually certify a new major version some time after the community release, so confirm when your provider supports PostgreSQL 19 before you fix a date.

How should teams plan the upgrade?

  1. Inventory versions. List every PostgreSQL cluster and replica, and flag anything still on 14 as the first priority.
  2. Check removed and changed settings. Search configs for RADIUS, MD5 roles, io_workers and JIT-dependent queries.
  3. Test on a production copy. Run pg_upgrade or a logical-replication cutover on a clone, then compare query plans and latency for your top queries.
  4. Pick a target per cluster. Move from 14 to 18 now if you need a low-risk step, or go straight to 19 where REPACK or replica reads justify it.
  5. Plan the rollback. Keep the old cluster or a verified backup until the new one has run under real traffic.

Frequently asked questions

When was PostgreSQL 19 released?

The PostgreSQL Global Development Group released PostgreSQL 19 on November 19, 2026. It was originally expected in late September or early October; the project moved the date after LLM-assisted code review surfaced more defects than developers could fix in time.

Which features were pulled from PostgreSQL 19?

The Register reported that 8 to 10 significant features were reverted before release, including SQL Property Graph Queries (SQL/PGQ). They are expected to return in a later major version, so do not plan graph-query work around PostgreSQL 19.

Does REPACK replace VACUUM FULL and CLUSTER?

REPACK combines what VACUUM FULL and CLUSTER do. With the CONCURRENTLY option it reclaims space and can reorder a table without blocking reads or writes. VACUUM FULL and CLUSTER stay available for compatibility.

What defaults changed in PostgreSQL 19?

JIT compilation is now off by default, TOAST compression defaults to lz4 when the server is built with LZ4, RADIUS authentication is removed, MD5 password logins now raise a warning, and io_workers is replaced by io_min_workers and io_max_workers.

Is PostgreSQL 14 still supported?

No. PostgreSQL 14.25, shipped with the November 2026 update releases, was the final release of version 14. Databases still on 14 get no further security fixes and should be upgraded.

Sources

PostgreSQL Global Development Group — PostgreSQL 19 Released! (primary source, November 19, 2026)
PostgreSQL documentation — Release notes for PostgreSQL 19
Phoronix — PostgreSQL 19 Released With More Performance Improvements (November 19, 2026)
The Register — Postgres 19 delayed after AI reviewers find more bugs than humans can fix (October 16, 2026)
The Register — PostgreSQL patches 52 security holes, sends version 14 into retirement (November 17, 2026)