Postgres 14 Reaches End of Life on 12 November. On RDS, the Date That Bills You Is 1 March 2027.
Community support and managed-service support end on different days. Which one you are on decides your timeline.
Two end-of-life dates, and you are on only one of them
Postgres 14 reaches community end of life on 12 November 2026, according to the Amazon RDS for PostgreSQL release calendar. After that date the PostgreSQL project stops shipping fixes for the 14 line. That is the date most headlines quote, and for a self-hosted database it is the one that matters.
If your database runs on Amazon RDS, a second date applies. The same calendar lists 28 February 2027 as the end of RDS standard support for version 14. RDS keeps shipping 14.x minors until then, including the engine versions released after the community stops. The calendar is explicit that you can keep running a version past that date for a fee.
So the useful question is not "when does Postgres 14 die?" but "whose support am I relying on?". A team on self-managed Postgres in Docker has days. A team on RDS has about three and a half months before billing changes. A team on another managed service has whatever that provider published, and you should read that provider's own page rather than a third-party summary. Some secondary sources quote Cloud SQL and Azure figures with confidence; treat those as leads to verify, not as schedules.
| Milestone | Date | What changes |
|---|---|---|
| Community end of life | 12 November 2026 | Upstream stops releasing fixes for 14 |
| RDS end of standard support | 28 February 2027 | Last day of included support on RDS |
| Extended Support year 1 pricing starts | 1 March 2027 | Paid per-vCPU charge begins |
| Extended Support year 3 pricing starts | 1 March 2029 | Higher pricing tier begins |
| End of Extended Support | 28 February 2030 | No further RDS support for 14 |
What actually breaks on 13 November
Nothing breaks. Postgres 14 will boot, serve queries and replicate on 13 November exactly as it did on the 11th. That is why end-of-life dates get ignored: the failure is not an outage, it is a gap that opens the next time a vulnerability is disclosed against the 14 line and no upstream patch follows.
The risk is therefore asymmetric. Most weeks it costs you nothing. In the one week a serious CVE lands, you are either patching on a deadline you did not choose or explaining to a customer why a database holding their data runs unsupported software. Security questionnaires and audits tend to ask for supported versions of critical components, so an unsupported database also shows up as a finding.
Extended Support is a bridge, not a plan
Cloud providers sell extended support so that a hard dependency does not force a rushed upgrade. The mechanism is pricing: the calendar above shows a separate year 1 and year 3 pricing start, which signals that the charge is designed to rise. Check the current rate on the RDS pricing page for your instance sizes and region, then multiply by every vCPU you run on 14, including replicas.
Use extended support when something concrete blocks you, such as an extension with no build for a newer major version or a vendor product certified only on 14. If the blocker is "we have not scheduled it", the fee is paying for a calendar entry.
Build the inventory before choosing a target
The expensive part of a major-version upgrade is rarely the upgrade command. It is discovering what you forgot. Before picking a target version, list every instance on 14 and who owns it.
-- Run on each instance you can reach
SELECT version();
SHOW server_version_num; -- 14xxxx means a 14.x server
-- Extensions that need a build for the target major version
SELECT extname, extversion FROM pg_extension ORDER BY 1;# List RDS Postgres instances and their engine versions
aws rds describe-db-instances \
--query "DBInstances[?Engine=='postgres'].[DBInstanceIdentifier,EngineVersion,DBInstanceClass]" \
--output tableRun both. The SQL check catches self-hosted servers and anything outside your cloud account; the CLI check catches the instance somebody created for a one-off analysis and never deleted. Add the extension list to the ticket, because extension support is the most common reason a target version has to change.
Choosing the target version
Target the newest major version your extensions, drivers and ORMs support, because each year you stay on the older release shortens the support you get from the upgrade. The RDS calendar lists community end of life in November 2029 for 17 and November 2030 for 18, so both leave years of runway. Version 18 was released in November 2025; if your dependencies support it, it buys the longest window. If a required extension lags, 17 is a reasonable fallback.
Whichever you choose, rehearse on a restored copy of production, not on a fresh empty database. Depending on the upgrade path and target version, planner statistics may not carry over, so check the upgrade notes for your route and run ANALYZE afterwards if they do not. A query plan that changed shape is the typical surprise, and you want to find it on a copy.
A plan that fits in a week
- Day 1: inventory every instance, replica and extension; assign owners.
- Day 2: restore a recent production snapshot into a test environment.
- Day 3: upgrade the copy, run ANALYZE, replay representative traffic or your test suite.
- Day 4: compare slow-query logs and plan changes against the old version.
- Day 5: write the rollback plan and book the production window.
Self-hosted teams should start now and accept that they will run a few days or weeks unsupported. RDS teams have until 28 February 2027 to avoid the fee, but the maintenance window, the freeze periods and the holiday calendar all eat into that. Counting backwards from the end of February, the practical deadline is earlier than the printed one.
What to watch next
Five-year support windows mean a major version leaves community support every November. Version 15 follows in November 2027. Teams that build an annual upgrade habit pay a small cost each year; teams that skip one pay it all at once, with a fee on top.
Frequently asked questions
Related reading
Idempotency Keys for Agent Tool Calls Need a Third State: "We Don't Know"
Most idempotency advice for LLM agents stops at "hash the arguments and keep a ledger". The failure that costs money is the crash after the side effect and before the record.
.NET 8 and .NET 9 Lose Support on the Same Day, 10 November. Your Inventory Is the Hard Part.
Microsoft ends support for .NET 8 and .NET 9 on 10 November 2026. Changing TargetFramework is easy; finding the Dockerfiles, CI images, hosts and packages that still pin the old runtime is the real work.
Auth0 Rules and Hooks Are Read-Only. Which End-of-Life Date Applies Depends on Whose Page You Read.
Some trackers list 18 November 2026 as the day Auth0 Rules and Hooks stop. Auth0's docs give 30 April 2027 for Hooks and 31 January 2028 for Rules. Here is what to check and how to migrate.