From: Achilleas Mantzios <achill@matrix.gatewaynet.com>
To: pgsql-admin@lists.postgresql.org
Subject: Re: zero data loss recovery is possbile with pgbackrest tool?
Date: Thu, 20 Jun 2019 13:40:08 +0300
Message-ID: <62943f11-43f6-fd4b-6311-92b7fce90670@matrix.gatewaynet.com> (raw)
In-Reply-To: <20190620121251.418294c0@firost>
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>
<20190620121251.418294c0@firost>
On 20/6/19 1:12 μ.μ., Jehan-Guillaume (ioguix) de Rorthais wrote:
> On Thu, 20 Jun 2019 08:37:21 +0300
> Achilleas Mantzios <achill@matrix.gatewaynet.com> wrote:
>
>> On 20/6/19 12:40 π.μ., 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 not
>>> 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?
>> No tool that is based on archiving is able to do that. Incomplete wals don'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 depending
> on various circumstances and activity.
In this case we have more than one problem.
In such cases pg_wal will be quickly stuffed with unarchived files, this indicates another (usually network or some sudden unforeseen spike in activity) problem.
So yes, in such cases where we have a network or performance problem AND at the same time disaster in the cluster occurs, then it would be much more than 16MB, but then again we are having two
problems now to solve (DB loss + infrastructure).
>
>> If you want to minimize the effect of this, you
>> should set
>> https://www.postgresql.org/docs/11/runtime-config-wal.html#GUC-ARCHIVE-TIMEOUT
>> (|archive_timeout|) to a smaller value .
> This will still allow data loss.
I said minimize.
>
>
> If you don't want a secondary in sync, you might want to have a look at
> pg_receivewal.
>
> ++
>
>
--
Achilleas Mantzios
IT DEV Lead
IT DEPT
Dynacom Tankers Mgmt
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-admin@postgresql.org
Cc: achill@matrix.gatewaynet.com, pgsql-admin@lists.postgresql.org
Subject: Re: zero data loss recovery is possbile with pgbackrest tool?
In-Reply-To: <62943f11-43f6-fd4b-6311-92b7fce90670@matrix.gatewaynet.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