Date: Tue, 25 Jun 2019 11:01:06 -0500
From: Justin Pryzby <pryzby@telsasoft.com>
To: Benjamin Scherrey <scherrey@proteus-tech.com>
Cc: Hugh Ranalli <hugh@whtc.ca>, pgsql-performance@lists.postgresql.org
Subject: Re: Perplexing, regular decline in performance
Message-ID: <20190625160106.GD18602@telsasoft.com>
References: 
 <CAAhbUMNdqFhdsoTKfE2wPp70+Zcrjs_Zcp7uh-6dnREZnVd5XA@mail.gmail.com>
 <CACo3ShijfteC03wGhsMwTOw16u39f9ameO=4RoQGz8b4SsQbvg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CACo3ShijfteC03wGhsMwTOw16u39f9ameO=4RoQGz8b4SsQbvg@mail.gmail.com>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk

On Tue, Jun 25, 2019 at 10:55:22PM +0700, Benjamin Scherrey wrote:
> Have you done a VACUUM ANALYZE FULL on your database? This needs to be done
> periodically to inform the server of the statistics of how the data and
> relations are distributed across the database.

I think this is wrong.

VACUUM and ANALYZE are good, but normally happen automatically by autovacuum.

VACUUM FULL takes an exclusive lock on the table, and rewrites its data and
indices from scratch.  It's not normally necessary at all.  It's probably most
useful to recover from badly-bloated table if autovacuum didn't run often
enough, in which case the permanent solution is to change autovacuum settings
to be more aggressive.

Don't confuse with VACUUM FREEZE, which doesn't require exclusive lock, and
normally not necessary if autovacuum is working properly.

Justin