Neues vom PostgreSQL Planet

I compiled and uploaded zip files with latest orafce and plpgsql_check for PostgreSQL 17 and PostgreSQL 18 - I used Microsoft Visual C 2022.

Setup:

You’ve probably seen the incident pattern:

1. Postgres backends start disappearing.
2. dmesg / journalctl -k shows the kernel OOM killer reaping postgres.
3. Someone spots “out of memory” and reflexively recommends: “Increase work_mem.”

That recommendation is frequently backwards for OS OOM kills.

For years now pgAdmin builds on Windows have been code signed using a certificate installed on a build machine in AWS. That certificate is expiring, so I had to look for a new one (I no longer work at a company that supports Windows that can provide one), and much to my annoyance I found that requirements have changed and a secure physical device is required for key storage.

As AI capabilities continue to evolve and integrate more deeply into our applications, we’re faced with interesting architectural decisions about how to expose our data to large language models (LLMs). Two approaches that have gained significant traction are Retrieval Augmented Generation (RAG) servers (such as pgEdge RAG Server) and Model Context Protocol (MCP) servers (such as pgEdge Natural Language Agent).

As AI capabilities continue to evolve and integrate more deeply into our applications, we’re faced with interesting architectural decisions about how to expose our data to large language models (LLMs). Two approaches that have gained significant traction are Retrieval Augmented Generation (RAG) servers (such as pgEdge RAG Server) and Model Context Protocol (MCP) servers (such as pgEdge Natural Language Agent).

I like reproducible development. I also like short feedback loops. Combining both for pgrx was… educational. 🙂 In this post, I share the mistakes, the small pains, and the fixes I used to get a working VS Code dev container for a Rust project that builds PostgreSQL extensions with pgrx. If you’re writing extensions or using pgrx in a team, this will save you a few grey hairs.

TL;DR:

New post up on on the ClickHouse blog:

New post up on on the ClickHouse blog:

I wanted to optimize away parsing the key/value pairs from the [pg_clickhouse] pg_clickhouse.session_settings GUC for every query by pre-parsing it on assignment and assigning it to a separate variable. It took a few tries, as the GUC API requires quite specific memory allocation for extra data to work properly. It took me a few tries to land on a workable and correct solution.