High disk activity Postgres stats collector process, how to deal?

Question

PostgreSQL

Hello. I have a small project on Amazon's Free Instance, and over the last few months I've been getting bills for exceeding the disk quota. I started to figure out who is tormenting the disk when there are no active users on the site - it turned out that the postgres statistics collector process. This is how the statistics look like for about 12 hours:


320 megabytes were written without any payload to the database. Google gives a bunch of complaints since the 8th version of postgre, but there is no sensible solution anywhere. It is advised to move the statistics file into memory, disable autovacuum, and so on. Maybe someone else has come across this and can suggest a solution.

Answer 1

It is treated like this:
1) mount tmpfs somewhere, for example in path
2) write in postgresql.conf: stats_temp_directory = path
This completely solves your specific problem with extra disk I / O, but, as you rightly noted, does not solve the problem in principle .
I wonder if you have many bases on this server? Are there many tables in databases?