Postgres Weekly

Postgres 19 now due November 19

Wed, 14 Oct 2026 00:00:00 +0000

#669 — October 14, 2026

Postgres Weekly

pg_vectorsearch: A Partition-Based Alternative to HNSW — A vector search extension that uses PRISM, a vector index that scans compressed k-means partitions rather than walking a graph like HNSW, and works on existing pgvector columns. It's alpha and requires Postgres 18+, while pgvectorscale moves to maintenance mode.

Erik Nordström et al. (Tiger Data)

Agent Memory: Raw Retrieval Vs. Extracted Facts, Tested — Redis reran agent memory systems on LongMemEval. Published scores missed reproduced results by up to 28 points, and pairing raw conversation retrieval with extracted facts beat either approach alone. See the data, costs and methodology.

Redis sponsor

How Savepoints Throttled Our Postgres Queue — A technical look, complete with some rather nifty visualizations, into how adding more workers to a Postgres-based queue made it slower because of per-job SAVEPOINTs (and how claiming jobs with a lease fixed it).

Rory Malcolm (incident.io)

IN BRIEF:

⭐ Postgres 19's release has been pushed back to November 19, with an extra RC2 on November 12 to include many more recent fixes. The first release candidate still lands tomorrow (October 15).
⚠️ PgBouncer 1.26.1 fixes three CVEs, the worst letting a client who knows one user's password log in as a different user (even a superuser) when SCRAM auth is in use.

'The Reverts Will Continue Until Morale Improves' — A Postgres committer, whose own work was among the casualties, shares his take on why so many Postgres 19 features were reverted during beta.

Peter Eisentraut

📊 Meanwhile, Kaarel Moppel ran 800 hours of benchmarks on the Postgres 19 betas and found them a few percent faster than v18 on OLTP workloads.

Don't Bet on SELECT Order: Shuffle Your Rows and See What Folds — If your app or tests implicitly rely on rows returning in insertion order, that can cause many an annoying bug (I've been bitten myself!). The pg_disorder extension shuffles SELECT results lacking an ORDER BY to catch such bugs in testing.

Yaroslav Kurbatov (Evil Martians)

What Deleting Rows Does to a Table — A tidy explainer and demo covering a common misunderstanding. Deleting half a table's rows, then running VACUUM, left it at 137 MB on disk, with only VACUUM FULL reclaiming it.

Hollis Varden

📄 Why Setting archive_command to /bin/true Can Lose Data – The 'fix' for a full WAL disk that breaks recovery without warning. Shaun Thomas

📄 PgBouncer Pooling Modes in Practice – What transaction pooling does to your advisory locks. Oskar Dudycz

📄 PgBouncer vs. Amazon RDS Proxy: How the Two Poolers Differ Wood and Chiba (AWS)

📄 How TABLESAMPLE Picks Rows Radim Marek

📰 Classifieds

RDST's always-on agent watches every PG query, automatically sends you explained fixes, and verifies they worked once applied. Put an agent to work on your Postgres database. It's free.

⚡ Rebase: An open-source backend for Postgres. REST, a typed SDK, auth, realtime, storage, and an admin panel you don't have to build.

Try ClickHouse Managed Postgres: Up to 10x faster for I/O-heavy workloads on local NVMe, with native ClickHouse integration.

RELEASES AND CODE:

pg-delta: Diff Postgres Schemas Into Ordered Migrations — Supabase's new open source schema diff engine. Point it at two Postgres 14–18 databases and get an ordered migration covering RLS policies, grants and domains. It replaces the deprecated migra, but is still alpha.

Valleteau and Xie (Supabase)

asyncpg 0.32.0: Fast Postgres Client Library for Python/asyncio — The long-standing Python client library for use with the asyncio framework now supports Postgres 19, as well as the newly released Python 3.15.

MagicStack

Orafce: Oracle Compatibility Functions for Postgres — Implements some of the functions from Oracle that might be useful for users migrating to Postgres. We haven’t linked to this in a few years but it continues to get frequent releases.

Orafce Project

wal2json 2.7 – A logical decoding plugin that emits row changes as JSON, used by many change data capture setups. This first release in over two years adds Postgres 18 and 19 support.
PGQueuer 1.6 – A Python job queue library built on Postgres.
Squawk 2.68 – A linter for Postgres schema migrations.

Microsoft's new MCP server for Postgres

Wed, 7 Oct 2026 00:00:00 +0000

#668 — October 7, 2026

Web Version

Postgres Weekly

Query Plan Rewriting in Postgres — A tour of the Postgres 18 planner's query transformations, including constant folding, inlining, subquery flattening, and join elimination. It's also useful to learn about the cases it won't rewrite, some of which can trip up experienced users (though Postgres 19 adds NOT IN rewriting when NULLs can be ruled out).

Phil Eaton

Govern EDB Postgres at Scale with Red Hat Ansible — Managing users, privileges, schemas, extensions and pg_hba by hand across a big Postgres fleet? See how edb.epas_postgresql, a Red Hat-certified Ansible Collection, enforces consistency at scale. Live with Red Hat and EDB, Oct 15.

EDB sponsor

Query Plan Hints in Postgres 19 — It may be “a new Postgres feature you will probably never need” but pg_plan_advice and pg_stash_advice are worth having on your radar for when you do need to steer the planner the right way. EXPLAIN (PLAN_ADVICE) will also let you see what's going on.

Elizabeth Garrett Christensen

IN BRIEF:

⭐ Postgres 19 RC1 is scheduled for October 15 and GA for October 29.
⏰ If you have B-tree indexes on char(n) columns, a REINDEX is coming your way soon. A newly committed, back-patched fix stops deduplication on bpchar indexes. It's expected to ship in November's minor releases.
Aurora PostgreSQL now has DuckDB embedded to query Iceberg and Parquet in S3 as foreign tables, a month after the DuckDB team joined AWS. Franck Pachot confirms it's DuckDB underneath by comparing plans with pg_duckdb.
Supabase is acquiring Turso, the company behind a Rust rewrite of SQLite. Supabase says it'll keep building around Postgres, with Turso's tech powering on-demand databases for AI agents.
PlanetScale now offers dedicated read replicas for Postgres.
Heroku Postgres Advanced is now generally available.

'PlanetScale Released Text Search and We Have a Lot to Say' — PlanetScale recently unveiled TIN, a full-text search extension, with benchmarks showing it to be much faster in queries than the open source ParadeDB. So ParadeDB got to work, and say they've “closed the gap”.

Ming Ying (ParadeDB)

Upgrading a 56TB Database by Mostly Not Moving It — Facing downtime due to a mandatory reboot deadline on Aurora, Trigger's engineers moved just 2% of their data to PlanetScale, leaving the rest behind. A candid postmortem involving a pgcopydb patch and a failed first cutover.

Daniel Sutton (Trigger)

Running Apache Cloudberry as Postgres Extensions — An experiment to run Apache Cloudberry (an open source descendant of Greenplum, the Postgres-based parallel data warehouse) as a set of extensions on a lightly patched Postgres 19. It surfaces some interesting limitations to Postgres' extension mechanism.

Igor Suhorukov

📄 What is Direct I/O, and Why Does ClickHouse Managed Postgres Use It for Backups? Kaushik Iska (ClickHouse)

📄 Resolving Replication Lag with Heartbeat Tables in CDC Scenarios Luo and Ooi (Amazon)

📰 Classifieds

Columnar storage without leaving Postgres: TimescaleDB compresses big tables up to 95% and speeds up scans. Get $1000 credit.

Visual editors for triggers and functions. A PL/pgSQL debugger that needs no extensions. Try dbForge Studio for PostgreSQL 2026.2 free.

⏱️ Postgres ran a prepared statement fast exactly five times. The sixth was 746% slower. Catch the flip in your query. Now in preview.

RELEASES AND CODE:

🤖 Microsoft Unveils an MCP Server for Postgres — They’re far from the first, but it’s interesting to see Microsoft step into this area with a Rust-powered (but installed via npm) MCP server to allow AI coding agents to work with Postgres, whether you’re on Azure or not.

Aditi Gupta (Microsoft)

💡 Microsoft also has a repo of 'Postgres skills' (which you can read here) covering both general Postgres usage and Azure platform specifics.

📊 pgwatch 6.0: Prometheus Sources, Rebuilt Dashboards, and a Sturdier Collector — The popular Postgres metrics collector and Grafana visualizer (demo) adds Prometheus exporters as sources, rebuilds its dashboards on Grafana 13, and copes better with databases that stop responding.

CYBERTEC PostgreSQL

pgx-bm25 1.0: A Pure C BM25 Index Extension — A new PostgreSQL-licensed BM25 ranked full-text search extension whose indexes live in the index relation's pages so WAL logging, replication, and VACUUM all work as usual with no separate engine or runtime to manage.

PGX Inc.

⚠️ pgBackRest 2.59.3 – Fixes a rare weak encryption subkey bug that upgrading alone won't repair, so check the notice if your repo is encrypted.
pg_textsearch 1.5 – Tiger Data's BM25 ranked text search extension. Adds boolean and phrase filtering via tsquery, plus Postgres 19 support.
⚠️ pgvector 0.8.7 – The ubiquitous vector search extension fixes a bug where creating an IVFFlat index could lead to arbitrary code execution.
Debezium 3.7 – The popular change data capture platform. Can now take incremental snapshots from read-only replicas and has better support for Postgres' geometric column types.
pg_ivm 1.16 – The incremental view maintenance extension adds Postgres 19 support.

Can REPACK CONCURRENTLY replace pg_repack?

Wed, 30 Sep 2026 00:00:00 +0000

#667 — September 30, 2026

Web Version

Postgres Weekly

Postgres 19 Beta 4 Released — Confirming recent developments, the latest beta is largely about what’s been removed, although numerous fixes have also made it in. A release candidate is due in the next couple of weeks, with the final release possibly in October if things go well.

PostgreSQL Global Development Group

💡 What's gone? SQL/PGQ (property graph) support, online enabling/disabling of checksums, temporal updates/deletes with FOR PORTION OF, MERGE/SPLIT PARTITIONS, and the pg_get_role_ddl(), pg_get_tablespace_ddl(), and pg_get_database_ddl() functions.

SELECT count(*) Shouldn't Take Eight Seconds — A count or a rollup over millions of rows crawls, so you add indexes, a read replica, a nightly job. TimescaleDB extends the Postgres you already run so those queries return in milliseconds on live data. No warehouse. Get $1000 credit to start.

Tiger Data (creators of TimescaleDB) sponsor

What REPACK (CONCURRENTLY) Costs While It Runs — A measurement of Postgres 19's new online table rewrite against pg_repack and pg_squeeze. REPACK was fastest, but has some downsides to consider.

Radim Marek

💡 REPACK (CONCURRENTLY) is getting a lot of attention on pgsql-hackers this week too. A bug that could lose updates was fixed, and Radim's own testing found a hard limit on how many concurrent changes a repack can absorb.

IN BRIEF:

⚠️ PgBouncer 1.26 addresses three CVEs, including two DoS bugs that unauthenticated clients can trigger. It also now tracks search_path and default_transaction_read_only by default.
Devrim Gunduz reports that Postgres's official Yum and Zypper repositories are now easier to navigate on the Web. He's also begun packaging Rust extensions there, starting with PostgreSQL Anonymizer.
Google has rounded up its upstream Postgres work so far this year.
Amazon Aurora now supports Postgres 18.6, 17.11, 16.15, 15.19 and 14.24.

Can Your Postgres Survive a Bad Query? — ClickHouse compares what happens on four Postgres cloud providers (itself included) when queries eat all the RAM, and explains why setting work_mem doesn’t prevent it.

Kevin Biju (ClickHouse)

▶ Andres Freund on Where Linux Helps and Hurts Postgres — A candid talk from this year’s Kernel Recipes conference, aimed at kernel developers, covering topics like io_uring, huge pages, futexes, and atomic writes.

Andres Freund

📄 PgQ and PgQue: The Queue Engines You Might Need – Spotting bloated task tables and moving them to a purpose-built queue. Jobin Augustine (Percona)

📄 Fixing One Function's Disk Spills with ALTER FUNCTION … SET work_mem – An API slowdown was traced to a function causing 150GB of temporary-file writes a day, but a function-local setting tidied up the mess. Dinesh Kumar (CortexRidge)

📄 When to Choose x86 or ARM for Postgres – Some pros and cons, if you're able to make the choice. Ahmed Darwich (PlanetScale)

📄 pgBackRest and Failover: Why archive_mode Matters Stefan Fercot

📰 Classifieds

🎟️ Join us at Supabase Select and learn how we’re building the next generation operating system for Postgres with Multigres. Apply to attend with 25% off: SUPAWEEKLY.

🛠️ Tools and Releases

Safe Not Safe: Is My Migration Safe? — A fun side project from a former lead of Cloudflare's Postgres platform team, who spent years reviewing other teams' migrations. Paste in a SQL migration and it flags common problems, parsing entirely in the browser using Postgres's own SQL parser via WASM.

Vignesh Ravichandran

Kafgres 0.2: 700 MB/s of Kafka Throughput on Postgres — We featured this Kafka-compatible broker that runs inside Postgres a few weeks ago. Its author has since taken it from 113 MB/s to 700 MB/s write throughput on a single box.

Raynor Elgie

Pgpool-II 4.7.3, 4.6.8, 4.5.13, 4.4.18, 4.3.21 – A raft of security fixes to fix seven CVEs, one allowing clients using certificate authentication to log in as another user without a password.
PgDog 0.1.60 – Rust-powered pooler and sharding proxy. Sharded SELECT queries are now handled 40% faster.
pg_raw_parse – PgDog's Rust bindings to the Postgres parser with 20–60x faster parsing than pg_query.rs.
pgBackRest 2.59.2 – The backup/restore tool adds Postgres 19 beta 4 support.
PLHaskell 6.0 – Write Haskell in SQL code. Now supports Range types.

✨ And finally…

'We Ported the Original Doom to SQL' — Not quite Postgres, but a fun project from the team behind the Postgres-compatible CedarDB. It takes 1993's pioneering FPS and runs the logic and renderer in SQL, returning each frame as a bitmap for Python to show. Some neat architectural details (the renderer uses 89 CTEs!) and SQL queries here.

Lukas Vogel (CedarDB)

Postgres 19's bumpy road to release

Wed, 23 Sep 2026 00:00:00 +0000

#666 — September 23, 2026

Web Version

Postgres Weekly

Looking Forward to Postgres 19: Epilogue — Postgres 19 is having an eventful beta, with several headline features pulled and no release date. Shaun digs into why via the pgsql-hackers mailing list, and suggests AI-assisted review is surfacing bugs faster than committers can handle them. For the good news, Elizabeth Garrett Christensen lists what's still shipping, like parallel autovacuum and ON CONFLICT DO SELECT.

Shaun Thomas (pgEdge)

💡 Beta 4 is expected tomorrow (September 24). This draft release announcement lists the reverts, and a release candidate should follow in early October.

Analytics Shouldn't Mean a Second Database — A dashboard query slows as data grows, and the fix is usually a warehouse plus a pipeline: sync jobs, lag, stale copies. TimescaleDB extends the Postgres you already run, so analytics stay fast on live data. Get $1000 credit to start

Tiger Data (creators of TimescaleDB) sponsor

In Brief

🔎 PlanetScale has launched TIN, a fast BM25 full-text search index for Postgres. It's closed source and PlanetScale-only, though they do offer Lead, an open source "slow but correct substitute" for testing TIN queries locally.
🎧 Over on Talking Postgres, Claire Giordano spoke with Peter Eisentraut about his 25 years of contributing to Postgres, which began with psql tab completion.
🇨🇦 PGConf.dev 2027 takes place May 11–14 in Montréal. Registration is open. It's always popular and you can watch 📺 this year's talks here.
Google has previewed native BM25 search in AlloyDB and Cloud SQL, built atop the pg_textsearch extension.

⚙️ Operations & Performance

pgBackRest Compression: How Much CPU is a Smaller Backup Worth? — A benchmark of pgBackRest's compression algorithms and levels to find where extra CPU stops paying off in smaller backups. Spoiler: "Zstandard at a low level is the sweet spot".

Agustín Gallego (Percona)

Ten Years of Postgres Logical Replication — A tour of what each Postgres release (from 10 to 19) added to logical replication, shown through building a multi-server write setup using only core features.

Dimitri Fontaine

When Deleting Old Rows Costs More Than the Work — Some lessons for anyone with big tables where bulk deletes can cost more than the writes. DBOS's engineers share some fixes, like replacing cascading deletes with batched deletes and running manual vacuums.

Kraft and Li (DBOS)

📄 Where Postgres Stores Row Locks – Every row lock ends up as a write to the heap page, as revealed by pageinspect. Radim Marek

🛠️ Tools and Releases

pgcli 4.7: The Friendlier psql Now Works in Scripts Too — A popular alternative to psql for interactive work, with context-aware autocompletion, syntax highlighting, and named queries. v4.7 makes it more useful for scripting, with psql-style -c, -f and -t flags, plus -y to skip destructive command prompts.

the dbcli project

pgwrh 1.0 Alpha: Read Scaling with Sharded Replicas — An experimental extension that shards a partitioned table across read replicas, with each holding only part of the data and fetching the rest from peers. Not for production!

Michał Kłeczek

pgColumnar 1.0 Alpha 4: A Columnar Table Access Method — A columnar storage option for Postgres tables. This alpha can lay tables out along a Hilbert curve, so range filters read up to 2x fewer row groups than with Z-order.

Joshua D. Drake (Command Prompt)

TimescaleDB 2.30 – Tiger Data's time-series extension makes ORDER BY ... LIMIT queries fast even on hypertables with huge numbers of chunks.
PostgreSQL Anonymizer 3.2 – Dalibo's data masking extension adds faster pseudonymization functions and includes fixes for 3 critical vulnerabilities.
pgAdmin 4 v9.18 – The popular Postgres management tool.

📰 Classifieds

🚀 Triple your Postgres productivity. pgEdge AI DBA Workbench: AI monitoring & diagnostics for ANY Postgres v14+. Open source.

🎟️ Join us at Supabase Select on October 2. Learn from the industry’s best builders and meet people building developer tools. Apply to attend with 25% off: SUPAWEEKLY.

And Finally…

🤖 Teaching a 4B Model to Write 81% Faster Query Plans — As a sabbatical project at the Recurse Center, a developer used reinforcement learning to train a small model to write pg_hint_plan hints for join-heavy benchmark queries (and, on its best runs, it beat Postgres's own plans!). An intriguing read.

Rohan Bansal