Subject: upgrades and streaming replication
From: Charles Sprickman <spork@biglist.com>
Content-Type: text/plain; charset=us-ascii
Message-Id: <BF48DCBA-CA71-4BF2-94CB-D95FD6909D6C@biglist.com>
Date: Tue, 10 Sep 2013 23:13:24 -0400
To: pgsql-admin@postgresql.org
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (Apple Message framework v1085)
Precedence: bulk
Sender: pgsql-admin-owner@postgresql.org

I just wanted to check if there are any updates on the preferred way to upg=
rade a group of servers that are using streaming replication with minimal d=
owntime.

Mailing list messages from a year or two ago pretty much state the followin=
g:

* No current plans to come up with some new mechanism to automate this proc=
ess, downtime is assumed.
* If you want no downtime, use one of the userspace replication solutions.
* Best option is to upgrade the master with pg_upgrade, then upgrade the sl=
aves and re-copy your data.
* While it's tempting to think you could just run pg_upgrade on all hosts t=
o avoid the copy step, that will not work as something in the system catalo=
g changes which will break replication.
* No good options to speed up the slave copy - rsync will either use timest=
amps to figure out if a file changed (and streaming skews that), or checksu=
mming (and checksumming is slow).

Is that still pretty much the situation today with 9.2 and 9.3?

Any good guides or BCPs for the standard procedure that emphasize limiting =
downtime and giving a good rollback path?

Thanks,

Charles

--=20
Sent via pgsql-admin mailing list (pgsql-admin@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-admin