Date: Thu, 20 Jun 2019 12:12:51 +0200
From: "Jehan-Guillaume (ioguix) de Rorthais" <ioguix@free.fr>
To: Achilleas Mantzios <achill@matrix.gatewaynet.com>
Cc: pgsql-admin@lists.postgresql.org
Subject: Re: zero data loss recovery is possbile with pgbackrest tool?
Message-ID: <20190620121251.418294c0@firost>
In-Reply-To: <a1eda6cc-cbe9-4859-52e8-b73fe7dd097c@matrix.gatewaynet.com>
References: 
 <CA+M0sHHPcx7n-n+OSL4QZ1wm4-qWNEzktCmrFmau2earuZRVhA@mail.gmail.com>
	<e158d1c5-c0a8-24d0-ce74-7706782a8391@gmail.com>
	<E776AABE-8CB4-4078-B9BA-40688E919AF3@crazybean.net>
	<CA+M0sHEamU3hs5b6FJSkRLG5yxhzrZyLCF2fNmxtXzjRzzhcuA@mail.gmail.com>
	<a1eda6cc-cbe9-4859-52e8-b73fe7dd097c@matrix.gatewaynet.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, 20 Jun 2019 08:37:21 +0300
Achilleas Mantzios <achill@matrix.gatewaynet.com> wrote:

> On 20/6/19 12:40 =CF=80.=CE=BC., Pavan Kumar wrote:
> > Hello Rui,
> >
> > yes those are committed transaction.
> >
> > I have tested two cases
> >
> > 1. complete lost. (means complete data directory, pg_wal direcoty)
> > 2. lost few directories (deleted few directories , except pg_wal).
> >
> > both cases I do see latest data in the current wal is lost.
> >
> > when I check the recovery process , wal archives are copying from
> > pgbackrest repository to pg_wal directory. however pgbackrest tool is n=
ot
> > copying current wal log at all. it is only copying archived wal's/
> >
> > is there any way to copy current wal log with pgbackrest tool?
>=20
> No tool that is based on archiving is able to do that. Incomplete wals do=
n't
> get archived till they are complete. So, you have the risk of losing
> (maximum) 16MB worth of data.

No, this could be more than 16MB. Archiving might be lagging as well depend=
ing
on various circumstances and activity.

> If you want to minimize the effect of this, you
> should set
> https://www.postgresql.org/docs/11/runtime-config-wal.html#GUC-ARCHIVE-TI=
MEOUT
> (|archive_timeout|) to a smaller value .

This will still allow data loss. No matter the value. Moreover, there no se=
nse
setting it bellow eg. 1min. If you want something smaller, you really want
replication.


If you don't want a secondary in sync, you might want to have a look at
pg_receivewal.

++