Date: Wed, 6 Mar 2019 11:05:14 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: Mariel Cherkassky <mariel.cherkassky@gmail.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: autovacuum just stop vacuuming specific table for 7 hours
Message-ID: <20190306170514.GY29584@telsasoft.com>
References: 
 <CA+t6e1k2wvczw2w3dG748uTi-jwXKbh1TY_1otjW0_wxJsTUAQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CA+t6e1k2wvczw2w3dG748uTi-jwXKbh1TY_1otjW0_wxJsTUAQ@mail.gmail.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Precedence: bulk

On Wed, Mar 06, 2019 at 06:47:21PM +0200, Mariel Cherkassky wrote:
> Those settings helped but the table still grey very much. I wrote a script
> that monitored some metadata about the table (pg_stat_all_tables,count(*)
> from orig and toasted table). I let the system monitor the table for a week
> and I found out the next info :

> Autovacuum was running great during the whole week and whenever it reached
> 10k records in the toasted table it started vacuuming the table. *However,
> The db grew dramatically during a period of 7 hours in a specific day. In
> those 7 hours the table contained more then 10k (and kept increasing) but
> the autovacuum didnt vacuum the table*. I saw that during those 7 hours
> autovacuum didnt run and as a result of that the table grew to its max
> size(the current size).

Does pg_stat_all_tables show that the table ought to have been vacuumed ?

SELECT * FROM pg_stat_sys_tables WHERE relid='pg_toast.pg_toast_123456'::regclass;

Compare with relpages, reltuple FROM pg_class

What postgres version ?

Justin