So, in the early days of the ARRP site I published a dump of the system tables, on a per-Redshift version basis. The code which produced the dump was ditzy - i

So, in the early days of the ARRP site I published a dump of the system tables, on a per-Redshift version basis. The code which produced the dump was ditzy - it just produced the pages which were uploaded. Back in about May 2022, I rewrote the backend so it would archive the contents of certain system tables (those needed to generate the pages about the system tables), and the existing pages fell into disuse, because they were a lot of manual work to maintain. I’ve just pushed the third test release of the Combobulator (AMI for Redshift management) to the first two test users, and it’s the first fully working version, which is great, and I’ve had a day or two off, just to kick back - and I’ve finally now written the code to dump HTML pages from the data I’ve been accumulating over the last seven or eight months. The current pages then are a first effort, which gets the pages up to date with the current release (and releases since May 2022). I’m already able to do something I wanted to do before, but couldn’t, which is show the size in kilobytes of the text for the SQL of views. Where I have the source data now in my own database, I can improve these pages over time. The dumps are useful in two particular ways. Firstly, the official docs are maintained by hand, not automated, and they’re ten years old - the system table pages list columns which no longer exist, and have not been updated to list columns which now exist, have incorrect data types and so on. My feeling (I’ve not run any code to compute a number) is they are now about 20% incorrect. Secondly, they provide easy and direct access to the SQL code for the system views. This is important because the system table docs are a veneer only, and so you find you need to look into the SQL to understand what you’re seeing. However, AWS have started censoring the SQL of more recent views - they’ve put code in to specifically block viewing the SQL. This isn’t very widespread yet, but it’s bloody stupid as there’s no up-side to it and interferes with the ability to system admins to manage their clusters.