Date: Thu, 25 Jan 2018 20:46:48 -0500
From: Stephen Frost <sfrost@snowman.net>
To: Michael Paquier <michael.paquier@gmail.com>
Cc: Azimuddin Mohammed <azimeiu@gmail.com>,
	Rakesh Kumar <rakeshkumar464@aol.com>, pgsql-admin@postgresql.org,
	pgsql-general@postgresql.org
Subject: Re: Sync replication
Message-ID: <20180126014648.GC2416@tamriel.snowman.net>
References: 
 <CAKUuFd4K7OC8s28=WtHQ5X=hpriqZD+JUCTyeHS3kj=jVq4Nkw@mail.gmail.com>
 <1612e8fd701-1722-18849@webjas-vae073.srv.aolmail.net>
 <CAKUuFd5GCiJip8q8QOMQL0bZUBDbGnXTNJVxoLzhKe0+nwZo_w@mail.gmail.com>
 <20180126012143.GB2416@tamriel.snowman.net>
 <20180126013738.GG17847@paquier.xyz>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="H2DV07OJo/LesPO+"
Content-Disposition: inline
In-Reply-To: <20180126013738.GG17847@paquier.xyz>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk


--H2DV07OJo/LesPO+
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Greetings Michael,

* Michael Paquier (michael.paquier@gmail.com) wrote:
> On Thu, Jan 25, 2018 at 08:21:43PM -0500, Stephen Frost wrote:
> > That's correct, which is why it's encouraged to have multiple replicas
> > configured when using synchronous replication.  In v10, it's possible to
> > specify how many synchronous replicas are required to have acknowledged
> > a given transaction before a COMMIT is returned to the applicaiton.
> > Please review the documentation here:
> >=20
> > https://www.postgresql.org/docs/current/static/warm-standby.html#SYNCHR=
ONOUS-REPLICATION
>=20
> Be careful about the performance impact if there is network latency
> between a primary and its standbys. One small correction to what Stephen
> says here. It is possible to define multiple synchronous standbys in
> v9.6. v10 has added the possibility to define quorum groups. Note that
> the grammar as been kept backward-compatible across versions.

Ah, yeah, I was thinking of quorum groups when writing that but you're
right that 9.6 added support for multiple synchronous standbys.  Older
versions also supported having multiple synchronous standbys configured
but were hard-wired to accept the commit with just one of them
acknowledging it (in a priorty order based on the order of them as
defined in synchronous_standby_names).

Thanks for the clarification!

Stephen

--H2DV07OJo/LesPO+
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
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=hLPu
-----END PGP SIGNATURE-----

--H2DV07OJo/LesPO+--