Message-ID: <45346.192.168.207.54.1427386574.squirrel@webmail.etecsa.cu>
In-Reply-To: <848F3721-4A21-4260-9640-8F698E3EE376@skogoglandskap.no>
References: <848F3721-4A21-4260-9640-8F698E3EE376@skogoglandskap.no>
Date: Thu, 26 Mar 2015 11:16:14 -0500 (GMT+5)
Subject: Re: postgres hot-standby questions.
From: "Gilberto Castillo" <gilberto.castillo@etecsa.cu>
To: "Graeme B. Bell" <grb@skogoglandskap.no>
Cc: "pgsql-admin@postgresql.org" <pgsql-admin@postgresql.org>,
 "Simon Riggs" <simon@2ndquadrant.com>
Reply-To: gilberto.castillo@etecsa.cu
User-Agent: SquirrelMail/1.4.8-5.el5_4.10
MIME-Version: 1.0
Importance: Normal
Content-Type: multipart/mixed; boundary="=tAFVpmAk72106=imx3.etecsa.cu="
Precedence: bulk
Sender: pgsql-admin-owner@postgresql.org

This is a multi-part message in MIME format.

--=tAFVpmAk72106=imx3.etecsa.cu=
Content-Disposition: inline
Content-Type: text/plain;charset=utf-8
Content-Transfer-Encoding: 8bit


>
> Hello everyone,
>
> Two questions, grateful for any feedback anyone can share. They relate to
> switchover between master and hot-standby.
>
>
> 1. What exactly is the behaviour of the master/primary during shutdown
> when it has a hot standby?
>
> https://wiki.postgresql.org/wiki/Hot_Standby
> http://www.postgresql.org/docs/current/static/hot-standby.html
>
> If I shut down the master, then afterwards when it is finished, I shut
> down the standby, will they contain identical logical databases & WAL
> records, e.g. assuming possible network failure?
>
> i.e.
> - Is the primary shutdown delayed until it has received notice from the
> standby that all WAL has been received?    ( my guess is: no)
> - Is the primary shutdown delayed until it has received notice from the
> standby that all WAL has been applied?    ( my guess is: no)
> - Can the primary generate any new WAL (e.g. checkpoint) or logical data
> changes during the shutdown process that might not be sent to the standby?
>   (my guess is: no)
>
> For example,
> https://vibhorkumar.wordpress.com/2014/06/30/switchoverswitchback-in-postgresql-9-3/
> "With this patch, the walsender process tries to send all outstanding WAL
> records to the standby in replication when the user shuts down the
> master."
> "tries"?
>
> That page also makes it seem like you have to manually check the WAL
> status. (section 2)
> Is there any way to make the primary's completion of shutdown
> automatically synchronous with completion of WAL on standby(s)?


What you used for you repltación? aplication u other.

>
> 2. Let's assume for the moment I have some crazy reason to prefer to avoid
> rsync where possible, such as its historical weird behaviour on HFS+
> filesystems or its present unreliable heuristic for syncing hard-links or
> its default behaviour of not using checksums. Can a controlled switchover
> (not failover) be built on wal_keep_segments alone?
>
> wal_keep_segments could be set to a fairly high number, maybe 10000
> (160GB) to allow standby catchup even after a day. Let's assume here than
> 99% of maintenance takes less than an hour, and that we're keeping rsync
> as a fallback for the worst case.
>
> e.g. here's the whole switchover process...
>
> shutdown A (Master)
> shutdown B (Standby)
> (A and B  should be identical in terms of WAL and logical data at this
> point).
> swap M/S configurations around
> start B (Master)
>
> do some work on A for an hour
> finish work on A
>
> start A (Standby)
> A catches up with B from wal_keep_segments.
>
>
> shutdown B (Master)
> shutdown A (Standby)
> (A and B  should be identical in terms of WAL and logical data at this
> point).
> swap M/S configurations around
> start A (Master)
> start B (Standby)
>

Remeber init the slave igual firt moment.

> Graeme Bell
>
>
>
>
> --
> Sent via pgsql-admin mailing list (pgsql-admin@postgresql.org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-admin---
> This message was processed by Kaspersky Mail Gateway 5.6.28/RELEASE
> running at host imx2.etecsa.cu
> Visit our web-site: <http://www.kaspersky.com>, <http://www.viruslist.com>
>


Saludos,
Gilberto Castillo
La Habana, Cuba

--=tAFVpmAk72106=imx3.etecsa.cu=
Content-Type: text/plain
Content-Disposition: inline

--- 
This message was processed by Kaspersky Mail Gateway 5.6.28/RELEASE running at host imx3.etecsa.cu
Visit our web-site: <http://www.kaspersky.com>, <http://www.viruslist.com>

--=tAFVpmAk72106=imx3.etecsa.cu=
Content-Type: text/plain
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
MIME-Version: 1.0


-- 
Sent via pgsql-admin mailing list (pgsql-admin@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-admin

--=tAFVpmAk72106=imx3.etecsa.cu=--