Message-Id: <487CC4BB.EE98.0025.0@wicourts.gov>
Date: Tue, 15 Jul 2008 15:38:11 -0500
From: "Kevin Grittner" <Kevin.Grittner@wicourts.gov>
To: "Lance Campbell" <lance@illinois.edu>,<pgsql-admin@postgresql.org>
Subject: Re: Backup and failover process
References: <B10E6810AC2A2F4EA7550D072CDE87601970FD@SAB-FENWICK.sab.uiuc.edu>
	<487C9741.EE98.0025.0@wicourts.gov>
	<B10E6810AC2A2F4EA7550D072CDE8760CDDE7F@SAB-FENWICK.sab.uiuc.edu>
In-Reply-To: <B10E6810AC2A2F4EA7550D072CDE8760CDDE7F@SAB-FENWICK.sab.uiuc.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

>>> "Campbell, Lance" <lance@illinois.edu> wrote: 
 
> What happens if you take an SQL snapshot of a database while
> creating WAL archives then later restore from that SQL snapshot and
> apply those WAL files?
 
What do you mean by "an SQL snapshot of a database"?  WAL files only
come into play for backup techniques which involve file copies, not
dumps done using SQL commands (like pg_dump).
 
> Will there be a problem if the transactions
> within the newest WAL file after the SQL snapshot was taken cause
> problems when they are applied?
 
Point In Time Recovery (PITR) backup techniques allow you to restore
to a specified point in time, so you could restore up to the point
immediately before the problem transactions.
 
> I would assume yes but I wanted to
> check if there was some type of timestamp that would prevent an
issue
> from occurring?
 
Take another look at this section -- in particular,
recovery_target_time.
 
I hope this helps.
 
-Kevin