Re: Clustered index to preserve data locality in a multitenant application?

[Ben Chobot <bench@xxxxxxxxxxxxxxx>]

On Aug 31, 2016, at 2:55 PM, Nicolas Grilly <nicolas@xxxxxxxxxxxxxxxx> wrote:

It looks like Instagram has been using pg_reorg (the ancestor of pg_repack) to keep all likes from the same user contiguous on disk, in order to minimize disk seeks.

http://instagram-engineering.tumblr.com/post/40781627982/handling-growth-with-postgres-5-tips-from

This is very similar to what I'm trying to achieve.

The article is 3 years old. I'd be curious to know if they still do that.

If what they did 3 years ago is similar to what you are trying to do today, who cares what they are doing today? (Besides using pg_repack instead of pg_reorg, of course.)

For what it's worth, we use pg_repack on a regular basis and it works exactly as advertised.