about synchronous_standby_names or sync replic

On Thu, Sep 22, 2011 at 2:34 PM, e-blokos <infos@e-blokos.com> wrote:

Hi Folks,

I wonder if it's possible to have a sync replication only from the the node
where the client
who update the DB and other node replicated asynchronously ?

example :
nodeCentral node1 node2 node3

user1  from node3 update db to nodeCentral (master) so nodeCentral
replicates synchronously node3 and later other nodes...

yep, put the names of all the nodes in synchronous_standby_names and
set synchronous_commit to local... then when someone sends a write
command to the server "set synchronous_commit to on" (of course, this
SET is a command you send to the server just like any other SQL
command)

don't forget to reset synchronous_commit or use set local, otherwise
you will let this node as the sync rep until disconnect

--
Jaime Casanova         www.2ndQuadrant.com
Professional PostgreSQL: Soporte 24x7 y capacitación

Thanks for your answer.
That's great,
so for later other nodes replication it
needs an asynchronous replic like pg_stat or even a good old rsync ?

Franck

----- Original Message -----
From: "Jaime Casanova" <jaime@2ndquadrant.com>
To: "e-blokos" <infos@e-blokos.com>
Cc: <pgsql-general@postgresql.org>
Sent: Thursday, September 22, 2011 11:38 PM
Subject: Re: [GENERAL] about synchronous_standby_names or sync replic

On Thu, Sep 22, 2011 at 2:34 PM, e-blokos <infos@e-blokos.com> wrote:

Hi Folks,

I wonder if it's possible to have a sync replication only from the the
node
where the client
who update the DB and other node replicated asynchronously ?

example :
nodeCentral node1 node2 node3

user1 from node3 update db to nodeCentral (master) so nodeCentral
replicates synchronously node3 and later other nodes...

yep, put the names of all the nodes in synchronous_standby_names and
set synchronous_commit to local... then when someone sends a write
command to the server "set synchronous_commit to on" (of course, this
SET is a command you send to the server just like any other SQL
command)

don't forget to reset synchronous_commit or use set local, otherwise
you will let this node as the sync rep until disconnect

--
Jaime Casanova www.2ndQuadrant.com
Professional PostgreSQL: Soporte 24x7 y capacitaci�n

Mmmh sorry for my ignorance,
I just learnt that PG 9.0 includes a built in replication.
My first PG install was in 1998, what a pleasure how
PG became today...

long life to PG community

Franck

----- Original Message -----
From: "Jaime Casanova" <jaime@2ndquadrant.com>
To: "e-blokos" <infos@e-blokos.com>
Cc: <pgsql-general@postgresql.org>
Sent: Thursday, September 22, 2011 11:38 PM
Subject: Re: [GENERAL] about synchronous_standby_names or sync replic

On Thu, Sep 22, 2011 at 2:34 PM, e-blokos <infos@e-blokos.com> wrote:

Hi Folks,

I wonder if it's possible to have a sync replication only from the the
node
where the client
who update the DB and other node replicated asynchronously ?

example :
nodeCentral node1 node2 node3

user1 from node3 update db to nodeCentral (master) so nodeCentral
replicates synchronously node3 and later other nodes...

yep, put the names of all the nodes in synchronous_standby_names and
set synchronous_commit to local... then when someone sends a write
command to the server "set synchronous_commit to on" (of course, this
SET is a command you send to the server just like any other SQL
command)

don't forget to reset synchronous_commit or use set local, otherwise
you will let this node as the sync rep until disconnect

--
Jaime Casanova www.2ndQuadrant.com
Professional PostgreSQL: Soporte 24x7 y capacitaci�n