On Tue, 4 Apr 2006, Thomas F. O'Connell wrote:
On Apr 4, 2006, at 4:05 PM, Marc G. Fournier wrote:
I'll be curious to hear stories of people using it for replication. The way I
interpret replication, there's an available database (even if read-only) on
both ends. With PITR/on-line backups, the way I understand it, there's no way
to provide availability to the recovery database because it's in a process of
continuous recovery. It qualifies as high availability in terms of a failover
solution, but the recovery database is not actually available until something
triggers it to recover, at which point any writing done to it causes it to
cease to be a replicant of the base database.
We started a project on it here: http://pgpitrha.projects.postgresql.org/
So far we have a working version of it in CVS which we are using at
travelpost.com. You're correct, the secondary system is only available after
you complete the PITR recovery, but it works well for us currently. Right now
we just make base backups 3 times daily and restore all the way from the base
when we cutover. The first thing we'll be changing is that methodology (i.e.
we'll be going to a continuous recovery methodology). Hopefully we'll get
some interest from more folks soon and get some good ideas flowing.
--
Jeff Frost, Owner <jeff@xxxxxxxxxxxxxxxxxxxxxx>
Frost Consulting, LLC http://www.frostconsultingllc.com/
Phone: 650-780-7908 FAX: 650-649-1954
