Category: PostgreSQL

TL;DR: Build a small derived Docker image that layers the pgvector extension onto the official postgres:17 image, deploy it with the Plesk Docker extension, persist data to the host, then CREATE EXTENSION vector; in your database. You’ll be up and running in minutes—with clean upgrades, backups, and security baked in.   What You’ll Build A PostgreSQL 17 container on a Plesk server (Ubuntu 22.04 assumed) The pgvector extension installed and available to all databases Persistent storage on the host at /var/lib/postgres/pg17-data A repeatable, pinned build (no “latest” surprises) Estimated time: 20–30 minutes   Prerequisites Plesk Obsidian with the Docker extension installed SSH access with sudo privileges Docker engine available on the host Basic familiarity with Plesk’s Docker UI (Add Container, environment variables, volumes, ports) Why a derived image? We pin the Postgres major version and the pgvector version for reproducibility and security. You can rebuild the image at any time

One of the most common headaches for engineers working with PostgreSQL is a version mismatch between the client tools (psql, pg_dump, pg_restore, etc.) installed on a host and the actual Postgres server they need to manage. For example: Your Ubuntu box ships with psql 12.17… but your database is running Postgres 16. You need to restore a dump created by pg_dump 16.1, but your host only has pg_restore 13.14. Installing the correct client from apt repositories can be a hassle, especially if you don’t want to add extra repos or upgrade the entire Postgres server stack. The good news? There’s a simple workaround: use the official Postgres Docker images as a source of up-to-date client binaries. Why client versions matter PostgreSQL is generally backwards compatible, but not always. Client tools expect certain dump formats and GUC parameters that may not exist in older/newer servers: A pg_restore from v12 may fail

In MySQL I always use “DESCRIBE TABLE” or “SHOW CREATE TABLE” to show the schema of a table but I always forget how to do this in PostgreSQL. The PostgreSQL equivalent is: