PostgreSQL

PostgreSQL 18: Add “–missing-stats-only” to vacuumdb

Loosing all the object statistics after a major version upgrade of PostgreSQL with pg_upgrade is one of the real paint points in PostgreSQL. Collecting/generating the statistics can take much longer than the actual upgrade which is quite painful. A(…)

PostgreSQL 18: More granular log_connections

Many of our customers enable log_connections because of auditing requirements. This is a simple boolean which is either turned on or off. Once this is enabled and active every new connection to a PostgreSQL database is(…)

pgvector, a guide for DBA – Part2 indexes

Introduction In the last blog post we did setup a LAB with a DVDRental database mixed with Netflix data on top which we created some embeddings hosted on the same tables thanks to pgvector. In this second part, we(…)

PostgreSQL: Unlogged tables and backups

Recently we faced the following situation and have been asked to analyze the issue: A dump was loaded into a primary instance of PostgreSQL, but the data did not reach the replica. Tables which have been added after the dump(…)

pg_mooncake: (another) Columnar storage for PostgreSQL

A very, very long time ago I’ve written a blog about cstore_fdw, which brings columnar storage to PostgreSQL. This is now part of Citus and does not anymore come as a separate extension. While this still can be(…)

PostgreSQL 18: Virtual generated columns

This is about another feature which was committed for PostgreSQL 18: Virtual generated columns. Generated columns are available already, but they need to be “STORED”. This means the result of an expression is stored to disk and the result is(…)

pgvector, a guide for DBA – Part1: LAB DEMO

Introduction The first release of pgvector occurred in April 2021. Since then a lot has changed and is still changing. Even though we could say that the level of maturity of AI stacks and usage is currently really amazing, things(…)

PostgreSQL 18: Introduce autovacuum_vacuum_max_threshold

Vacuum/Autovacuum is one of the critical parts of every PostgreSQL installation. When autovacuum is not configured properly for your workload you’ll suffer from bloat and performance issues sooner or later. Most of the installations we’ve seen run with(…)

PostgreSQL: Indexes and casting

This is a small reminder to be careful with casting one data type to another in your queries when you want to have an index access rather than a sequential scan. Here is a small example of what can happen:(…)

PostgreSQL 18: Per-relation cumulative statistics for [auto]vacuum and [auto]analyze

This is about another feature which will most likely show up in PostgreSQL 18 later this year. The statistic system is something which gets more and more details with almost every release of PostgreSQL, and PostgreSQL 18 will be no(…)