From: | Matthias Müller <mmblnfrn(at)gmail(dot)com> |
---|---|
To: | pgsql-general(at)postgresql(dot)org |
Cc: | Satoshi Nagayasu <satoshi(dot)nagayasu(at)gmail(dot)com>, Ben Carbery <ben(dot)carbery(at)gmail(dot)com> |
Subject: | Re: Understanding PG9.0 streaming replication feature |
Date: | 2010-12-22 00:10:22 |
Message-ID: | 201012220110.23128.mmblnfrn@gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
Hi Ben,
load balancing is not possible with the tools that are in the postgres
installation. There is no automatic switch-over to a slave if the master
fails. The trigger file needs to be created to make a slave to the master. This
is not done automaitcally by postgres, but should be done by a cluster
software (i.e. pacemaker).
If you can live without load balancing, read the page of Satoshi.
But the slaves can handle read requests (SELECT). Maybe this helps a little
bit.
Regards
Matthias
Am Mittwoch, 22. Dezember 2010, um 00:31:14 schrieb Satoshi Nagayasu:
> Hi Ben,
>
> On 2010/12/22 7:46, Ben Carbery wrote:
> > FYI, not looking for a detailed how to here.. I have read the manual
> > twice and just can't figure which sections are relevant. The manual
> > seems to be trying to cover all uses simultaneously which is always
> > going to get confusing :) For example do I need I need WAL archiving or
> > not?
>
> My blog entry would be a good entry point for you. :)
>
> 5 steps to implement a PostgreSQL replication system
> http://pgsnaga.blogspot.com/2010/05/5-steps-to-implement-postgresql.html
>
> It was written to be a guide for building a simple master-slave config.
>
> Please take a look, including the comments.
>
> Thanks,
>
> On 2010/12/22 7:46, Ben Carbery wrote:
> > FYI, not looking for a detailed how to here.. I have read the manual
> > twice and just can't figure which sections are relevant. The manual
> > seems to be trying to cover all uses simultaneously which is always
> > going to get confusing :) For example do I need I need WAL archiving or
> > not?
> >
> > On Tue, Dec 21, 2010 at 2:40 PM, Ben Carbery <ben(dot)carbery(at)gmail(dot)com
<mailto:ben(dot)carbery(at)gmail(dot)com>> wrote:
> > Hi,
> >
> > I am having some trouble trying to figure out how to configure this
> > particular scenario..
> >
> > I have a pair of pg servers that I want to put in a Master/Standby
> > configuration. Currently a script dumps the master db every hour,
> > copies it to the standby, restores, and restarts the server. The aim
> > is to replace the dumps/restores with streaming replication and
> > ensure the standby is always up to date.
> >
> > In this case writes are infrequent, but reads are constant, and I
> > only need high availability for reads. I would ideally like both
> > master and standby to be available simultaneously to allow
> > load-balancing. My confusion seems to be around the fact I don't
> > need failover - my applications will detect a master down and
> > immediately start using the standby, so there is no need to allow
> > writes on the standby, they will just wait for the master to be
> > available again - I am not sure what the minimum config needed for
> > this scenario is..
> >
> > cheers,
> >
> > Ben
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2010-12-22 00:11:58 | Re: libpq sendQuery -- getResult not returning until all queries complete |
Previous Message | Merlin Moncure | 2010-12-22 00:03:52 | Re: libpq sendQuery -- getResult not returning until all queries complete |