Neues vom PostgreSQL Planet

Read/write splitting is easy for single, autocommit statements. It gets hard the moment a client opens a transaction. Most proxies give up and send every statement inside a transaction to the primary, even the reads. This article is about a middle path we built into Polar Proxy: keep the transaction on the primary, but let its read statements borrow a replica connection.

GitLab has a public story about deleting the wrong thing. In January 2017 an engineer who was trying to restore replication wiped a data directory they believed was on the secondary. It was on the primary. The postmortem they published afterwards is still one of the better things written about a database outage.

It’s been exactly one month since I published my work porting Greenplum to PostgreSQL 19 and isolating the minimal PostgreSQL core patches needed to run this minimal fork. It all started as a way to extract its ORCA analytical query planner and turn PostgreSQL into a modern, vectorized, high-performance DBMS. And I pulled it off! After a month of development, the database keeps getting faster. I’m now working on a GPU version.

A financial firm hired me in 2010 to revise their Postgres cluster architecture. During my stint there, I learned a healthy respect for their event_log table. But perhaps “respect” is the wrong word here, as it was more of a deep loathing. This table, you see, was the canonical ledger of system events and consisted of several million rows.The high row count wasn’t a problem by itself. We kept a tight visibility window and archived frequently. Unfortunately for us, one of the columns was an unconstrained TEXT type that commonly held a large blob of XML.

Over the past decade, the timing of PostgreSQL major releases has become very predictable: no earlier than the middle of September, and no later than the middle of October. This year is going to be different. PostgreSQL 19 most likely won’t be released until November, and it’s not impossible that the project could decide to postpone the release further.

The PostGIS Team is pleased to release PostGIS 3.7.0! Best Served with PostgreSQL 19 RC 1 , GEOS 3.15.0 , postgis_tiger_geocoder 2025.2 , and address_standardizer.

If you store PHI on Supabase, you need a signed BAA and a project set to High Compliance. That takes care of the platform. It doesn't cover your RLS policies, your functions or your audit tables, and a HIPAA review will ask about those too.

Supabase's shared responsibility model says the same. It tells you to "always apply Row Level Security (RLS)" and leaves the access rules on sensitive tables to you.

PostgreSQL 16.4, 15.8, 14.13, 13.16 and 12.20 added a security parameter that only pg_dump uses. (I consult on PostgreSQL through PGX Inc..)

Agent memory has been in fashion ever since agents appeared: it fights hallucinations and lets an LLM base its answers and decisions on your own information. Whether it is better to query that information from tables or from a graph is often decided on faith, after reading posts on the internet, without experiments and measurements of your own.

After pg_dump and pg_restore, by default the planner statistics and the visibility map are missing. One query: 32.6 s before VACUUM ANALYZE, 23 ms after.

PostgreSQL's `restore_command` asks for WAL segments one at a time until the command fails—and that failure marks the archive's end, whether it's real or not. (I consult on PostgreSQL through PGX Inc..)

PostgreSQL's no pg_hba.conf entry for host, measured on 12 to 18 (not 15): the 'no encryption' tail, Docker's gateway address, hostssl, replication and reload.

When a PostgreSQL backend crashes, the postmaster can restart itself automatically or exit and wait for external intervention. (I consult on PostgreSQL through PGX Inc..)

SQL vs PL/pgSQL functions in PostgreSQL 16: why a SQL function is re-planned on every loop call, how to reduce plan time.

For years, installing many packages from the PostgreSQL RPM repository on RHEL and its rebuilds started with the same step: `dnf install epel-release`. PostGIS, pgBackRest, pgcli, patroni, pg_statviz, ora2pg and lots of extensions pulled at least one dependency from EPEL.

Over the last couple of weeks I removed most of those dependencies. On RHEL 8, 9 and 10, the GIS stack, the most popular extensions and almost all of our Python tools now install from the PGDG repositories and the OS repositories (BaseOS, AppStream and CRB) alone.

As promised during the session, I'm sharing the slides from the "Fostering Inclusivity with the PostgreSQL Europe Diversity Committee" presentation at the New PostgreSQL Community Members - Summit at PostgreSQL Conference Europe (PGConf.EU) 2026.

Most teams I work with have a backup dashboard, and most days it is green. The job ran; it finished; nothing failed. That green tick feels like safety, and it is the reason nobody thinks about backups again until the day they need one. Then something breaks, someone asks how long it will take to get the database back, and the room goes quiet. Nobody knows. The backups have been running for a year, and nobody has ever brought one back. 

Most database proxies that do read/write splitting get one thing wrong: after a client writes a row and immediately reads it back from a replica, the read can return stale data.

Every commit in PolarDB ends the same way: the WAL record must reach shared storage. On the way there, the bytes move from a WAL buffer, through the shared-storage daemon, to a device.

PostgreSQL 16's `reserved_connections` reserves slots for your monitoring and backup roles when the application exhausts the pool. (I consult on PostgreSQL through PGX Inc..)