From: MichaelDBA <MichaelDBA@sqlexec.com>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Michael Lewis <mlewis@entrata.com>
Cc: Laurenz Albe <laurenz.albe@cybertec.at>
Cc: Daulat Ram <Daulat.Ram@exponential.com>
Cc: pgsql-performance@lists.postgresql.org <pgsql-performance@lists.postgresql.org>
Subject: Re: Shared_buffers
Date: Tue, 12 Mar 2019 16:11:49 -0400
Message-ID: <5C881285.8030905@sqlexec.com> (raw)
In-Reply-To: <20190312201102.GI23597@telsasoft.com>
References: <BL0PR01MB51871A89E9255FFCD13CC3839B480@BL0PR01MB5187.prod.exchangelabs.com>
<56099ac31c7cc1a4e8a51963e2674e7fe592b86f.camel@cybertec.at>
<CAHOFxGqT24bb_G9Ej3jAwBDLizfK1dwb+rXo62mP4zjB6iDRpA@mail.gmail.com>
<5C88107F.7040902@sqlexec.com>
<20190312201102.GI23597@telsasoft.com>
Here's one cook article on using pg_buffercache...
https://www.keithf4.com/a-large-database-does-not-mean-large-shared_buffers/
Regards,
Michael Vitale
> Justin Pryzby <mailto:pryzby@telsasoft.com>
> Tuesday, March 12, 2019 4:11 PM
>
> I've tuned ~40 postgres instances, primarily using log_checkpoints and
> pg_stat_bgwriter, with custom RRD graphs. pg_buffercache does provide some
> valuable insights, and I know it's commonly suggested to check
> histogram of
> usagecounts, but I've never had any idea how to apply that to tune
> shared_buffers.
>
> Could you elaborate on what procedure you suggest ?
>
> Justin
> MichaelDBA <mailto:MichaelDBA@sqlexec.com>
> Tuesday, March 12, 2019 4:03 PM
> Set shared_buffers more accurately by using pg_buffercache extension
> and the related queries during high load times.
>
> Regards,
> Michael Vitale
>
>
> Michael Lewis <mailto:mlewis@entrata.com>
> Tuesday, March 12, 2019 3:23 PM
> On Tue, Mar 12, 2019 at 2:29 AM Laurenz Albe <laurenz.albe@cybertec.at
> <mailto:laurenz.albe@cybertec.at>> wrote:
>
> Daulat Ram wrote:
> > I want to know about the working and importance of
> shared_buffers in Postgresql?
> > is it similar to the oracle database buffer cache?
>
> Yes, exactly.
>
> The main difference is that PostgreSQL uses buffered I/O, while
> Oracle usually
> uses direct I/O.
>
> Usually you start with shared_buffers being the minimum of a
> quarter of the
> available RAM and 8 GB.
>
>
> Any good rule of thumb or write up about when shared buffers in excess
> of 8GBs makes sense (assuming system ram 64+ GBs perhaps)?
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-performance@postgresql.org
Cc: MichaelDBA@sqlexec.com, pryzby@telsasoft.com, mlewis@entrata.com, laurenz.albe@cybertec.at, Daulat.Ram@exponential.com, pgsql-performance@lists.postgresql.org
Subject: Re: Shared_buffers
In-Reply-To: <5C881285.8030905@sqlexec.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox