PostgreSQL 14 is scheduled to receive its last minor release on November 12, 2026, according to a community post published on dev.to and credited to since.dev. The author states that after that date the PostgreSQL project will stop shipping fixes for the branch, including security fixes. The claim is the author's reading of the project's versioning policy and release roadmap, not an independently verified announcement in the supplied material.

The post lays out the reasoning behind the date. PostgreSQL supports each major version for five years from its first release, then issues one final minor release and stops. Version 14 first shipped on September 30, 2021, which puts the end of its window in late 2026, and the author says the final release coincides with the next scheduled round of minor releases.

The practical consequence the author describes is narrower than a shutdown. Minor releases are the vehicle for bug fixes, security patches and data-corruption fixes, so once 14 stops receiving them, anything discovered afterward stays unfixed on that branch. The database itself keeps running; what ends is the supply of corrections.

The post also argues that the successor to target is version 18, not 19. It says PostgreSQL 19 reached its first release candidate on October 15, 2026 and is not yet on the supported list, while 18 is the newest supported major and therefore has the most years remaining. Readers are advised not to wait for 19.

Scale is offered as context rather than measurement: the author cites an approximate GitHub code-search count, taken November 2, 2026, of about 72,000 public files containing the string image: postgres:14. That figure is a search-result count reported by the author, not an audit of running deployments, and it says nothing about private repositories or managed instances.

For teams that want to find their own pins, the post suggests repository searches for image tags, engine_version settings and distribution package names. It also warns that an untagged postgres image resolves to whatever major version is newest when the container is created, so the author recommends pinning the major version that has actually been tested.

Managed database services are a separate case, and the author is explicit that the provider's calendar governs. The post states that Amazon RDS for PostgreSQL keeps 14 in standard support until February 28, 2027, then moves it to paid Extended Support until February 28, 2030, and that Google Cloud SQL, Azure Database for PostgreSQL and Aurora set their own schedules. On a managed service, the author says to use the provider's major-version upgrade tooling rather than running the local tools yourself, while keeping the same testing steps.

The first of three documented upgrade routes is pg_upgrade. The author describes installing 18 alongside 14, creating a new cluster with initdb, and running a check-only pass first, since the check flag validates both clusters without altering data. The real run requires stopping both servers and dropping that flag, and a link mode skips file copying at the cost of losing the ability to return to the old cluster once the new one starts.

Two pg_upgrade details in the post are easy to miss. PostgreSQL 18's initdb enables data checksums by default and pg_upgrade requires both clusters to match, so a 14 cluster without checksums needs the new one created with checksums off. The author also says to install the 18 build of every extension's shared library beforehand and not to run CREATE EXTENSION in the new cluster, because pg_upgrade copies those definitions across.

The second route is dump and restore, which the author calls the simplest and slowest. For production, the recommendation is pg_dumpall from the newer version, which the post says can read servers back to 9.2. The author warns against simply changing an image tag on a volume holding 14 data, because major versions change the data directory format and 18 cannot start on it.

A Docker-specific note accompanies that route: from 18 onward, the author says Docker's image documentation directs mounting the volume at /var/lib/postgresql rather than /var/lib/postgresql/data. The sample sequence for a disposable Compose database dumps, removes the volume, brings the stack up on the new image and restores, and the author cautions that the volume-removal step is destructive, so the dump should be checked first.

The third route is logical replication, which the author says works across major versions: build an 18 server, let it catch up from 14, then switch the application over. It requires wal_level set to logical on the 14 server, which needs a restart, followed by a publication on the old server and a subscription on the new one.

The post lists the restrictions that make this route riskier than it first appears. Schema does not replicate, so it must be copied separately with a schema-only dump and frozen until the switch. Sequences do not replicate, so they must be set to current values on 18 just before writes move, or inserts will collide. Large objects do not replicate. And tables need a replica identity, by default the primary key, for updates and deletes to replicate; without one, the author says those statements fail on the 14 server, which at that stage is still production.

Before switching, the author recommends pointing continuous integration at 18 first: change the service container image, run the full suite, and fix failures before touching production. The post then lists specific checks, including confirming every extension listed on 14 is available for 18, comparing query plans with EXPLAIN (ANALYZE, BUFFERS) on the slowest and most frequent queries, and updating connection drivers to releases that support 18.

Two compatibility items on that checklist trace back to releases earlier than 18, not to 18 itself. The author notes that 18 deprecates MD5 passwords, and that a warning is issued when one is set; roles that use MD5 should therefore move to SCRAM, and drivers must handle it. From version 15 onward, new databases do not let every user create objects in the public schema; an upgrade preserves old permissions, yet a fresh 18 database in CI or a new environment will lack them, so a non-owner role running migrations can fail.

The post also flags that version 15 renamed the backup start and stop functions, which matters for backup scripts, and directs readers to the migration section of each release note from 15 through 18, where breaking changes are listed. It notes that upgrading to 18 keeps most planner statistics but not those created with CREATE STATISTICS, and recommends analyzing every database afterward.

For this audience, the editorial reading is that the deadline is a maintenance decision rather than an outage. Freelancers and small teams maintaining client projects on 14 have a dated window to plan an upgrade, and the cheapest first step supported by the post is inventory: find where the version is pinned, then check whether a managed provider's date differs from the community one. The author's CI-first advice is a low-risk way to surface extension, driver and permission problems before production is touched.

What remains unknown is substantial. The supplied material contains no primary PostgreSQL project announcement, no confirmation of the November 12 date beyond the author's citation of the roadmap, and no independent verification of the GitHub count. The post does not state how long each upgrade method takes, does not cover every managed provider's schedule, and does not claim that any of these procedures were tested by the author on a specific deployment. Readers should treat the dates and provider terms as claims to confirm against the project's and provider's own documentation before scheduling work.