Neues vom PostgreSQL Planet

Tom Lane deleted `array_nulls` in PostgreSQL 20, but `pg_dump` never set it, so restores into databases with it disabled silently corrupt text arrays. (I consult on PostgreSQL through PGX Inc..)

Most slow queries I have looked at are not doing anything clever. They are doing extra work that nobody asked for, because one word in the SQL told PostgreSQL to.

I took five of those words and measured each one against the version without it, on the same table. The biggest gap was UNION: 13.6 seconds, against 0.68 seconds for UNION ALL on the same two halves of the table.

Setup

One table, u, with 2,000,000 rows: a bigint identity key, an email, a status, a timestamptz and an md5 note.

Yay, “That time of the year”™ again soon 🎉 This time maybe with a bit more fuzz / excitement / anxiety as usual - due to various obstacles surrounding the release as detailed in a few recent posts by various people. Not going into the “Is AI to blame or...

On September 30, AWS announced that Aurora PostgreSQL can now query Apache Iceberg and Parquet data directly. That brings the number of things a vendor might mean by “Postgres reads Iceberg” to at least four, and they have about as much in common as the four things a vendor might mean by “serverl… (I consult on PostgreSQL through PGX Inc..)

The numeric type has existed in Postgres for more than 25 years. Over that time, it has undergone many optimisations and refinements.

PostgreSQL 18's new `num_os_semaphores` preset reports how many OS semaphores the server needs—a formula that drifted wrong for twelve years until the… (I consult on PostgreSQL through PGX Inc..)

Last week I ported Apache Cloudberry to PostgreSQL 19 as a set of extensions and described where the port falls behind the original fork. On ClickBench, columnar engines beat it by an order of magnitude. The culprit isn’t the MPP implementation but PostgreSQL’s executor, which works on rows. Each row travels up the node tree on its own, every value goes through fmgr, and the cluster’s segments merely add more processors running the very same loop over table rows.

Dive deeper into connection pooling with five more PostgreSQL poolers: Odyssey, pgagroal, Supavisor, PgDoorman, and ProxySQL, plus where your cloud already… (I consult on PostgreSQL through PGX Inc..)

This is part 2 of a series about Postgres logical replication use-cases, and about how the feature set has evolved over the past ten years and ten releases, one architecture at a time. Part 1 built a hub-and-workers system for write scalability and has the table of what each release added, which this post assumes. Part 3 covers a zero-downtime major upgrade.

The `notify_buffers` GUC tunes a 128kB cache for PostgreSQL's LISTEN/NOTIFY queue—and almost never matters. (I consult on PostgreSQL through PGX Inc..)

Last week I compared UUIDv4 and UUIDv7 as primary keys. There is a third case I left out: the key is a UUID, but the column is text. Some ORMs and JSON-first codebases end up there because a string is the easy path.

So I kept the same test and changed only the column type. 5 million rows, UUIDv7 and UUIDv4, each stored as uuid and as text. Short answer: the text index was 88% bigger, one level deeper, and every random lookup read one more page.

Setup

Same as last time. One table per run:

Two EXPLAIN ANALYZE runs on statistically identical data produced opposite plans, 73 heap blocks against 3,449. The margin was 1.5%. Here's the mechanism, measured with ExoBench, and the index shape that takes the coin out of the planner's hand.

My tool tells you things like “this index cuts the query’s cost by 91%”. Before releasing it, I wanted to know what that number is actually worth.

The tool is PgLens, a Postgres index advisor I’m building. Like several other advisors, it checks each suggestion with HypoPG: create a hypothetical index, run EXPLAIN again, and keep the suggestion if the planner’s estimated cost drops by at least 15%.

PostgreSQL connection poolers lie to both sides of the wire: your app thinks it has thousands of connections while the database sees dozens. (I consult on PostgreSQL through PGX Inc..)

pgvector 0.8.7 fixes CVE-2026-103484: a database user who can create an IVFFlat index can write out of bounds in the backend, which can lead to arbitrary code execution. Every version through 0.8.6 is affected. Upgrade. That part is simple. Two other parts are not: who can actually reach the bug,… (I consult on PostgreSQL through PGX Inc..)

On 29 September, the PostgreSQL Usergroup NL met, organized by Gerard Zuidweg and Sebastiaan Mannem

Speakers:

• Peter Lengkeek
• Ellert Van Koperen
• Jan Karremans

The Postgres US Summit 2026 was held from 30 September to 2 October.

Organizers:

When one long transaction holds row locks on thousands of parents, multixact cache misses can tank insert throughput. (I consult on PostgreSQL through PGX Inc..)

`min_wal_size` sets a floor under WAL recycling, not a retention policy—confusing many operators who lose standbys when the floor causes segments to be… (I consult on PostgreSQL through PGX Inc..)

List-partitioning by tenant under row-level security, measured in PostgreSQL 17 and 18: 100, 1,000 and 3,000 partitions, planning, locks, adding a tenant.

Parallel scans need a minimum table or index size to be considered—but that minimum also sets the bottom rung of the ladder the planner climbs to decide how…