Re: Recovery Test Framework

On Mon, 2009-01-12 at 20:52 -0500, Robert Haas wrote:

> I think the
> base backup should be integrated into the mechanism as well. I want
> to just be able to configure the master and slave for replication,
> fire up the slave, and walk away. Without that, I agree that it's
> likely to be too cumbersome for any actual use.

If you want integrated base backup, I would ask that we add it in the
next release and make it optional. It isn't necessary for sync rep and
is not a reason to slip that project; it's just icing. Many users have
been doing base backups for 2 releases now and I've never had a single
comment that it is cumbersome. The reverse actually, people say it is
flexible.

The flexibility of the current system is important for another reason.
"Integrated" will definitely mean single threaded because you just
aren't going to make it so complex. Single threaded has huge negative
implications in practice and we should not forget that the ability to do
a multi-threaded base backup is a critical user requirement.

Slony provides automated "base backup" transfer but does so using only a
single thread. So large databases take a long time to transfer. I have
spent time this year working with Jan, following up on two separate
ideas to improve this. The last one of those was looking at ways to
allow Slony to start via a base backup, just as warm standby allows.

Many users have found Warm Standby simple to configure and one part of
that is the ability to use parallel utilities to achieve the base
backup.

Restricting the way bulk copying happens only prevents innovative
solutions such as split mirrors, snapshot copies or whatever. We cannot
judge what the best way to ship giga or even terabytes of data to
another site will be for any user.

--
Simon Riggs www.2ndQuadrant.com
PostgreSQL Training, Services and Support