Date: Thu, 17 Nov 2005 18:25:11 +0100
From: "Steinar H. Gunderson" <sgunderson@bigfoot.com>
To: pgsql-performance@postgresql.org
Subject: Re: Help speeding up delete
Message-ID: <20051117172511.GA28121@uio.no>
Mail-Followup-To: pgsql-performance@postgresql.org
References: <43790A99.9050603@noao.edu> <4162.1132011763@sss.pgh.pa.us>
	<1132171206.4959.60.camel@localhost.localdomain>
	<437BDF9A.4060700@familyhealth.com.au>
	<20051117111927.GB26459@uio.no>
	<437C9CBE.8020907@familyhealth.com.au>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <437C9CBE.8020907@familyhealth.com.au>
User-Agent: Mutt/1.5.11

On Thu, Nov 17, 2005 at 11:07:42PM +0800, Christopher Kings-Lynne wrote:
>> Isn't your distribution supposed to do this for you? Mine does these 
>> days...
> A distribution that tries to automatically do a major postgresql update 
> is doomed to fail - spectacularly...

Automatically? Well, you can install the two versions side-by-side, and do
pg_upgradecluster, which ports your configuration to the new version and does
a pg_dump between the two versions; exactly what a system administrator would
do. Of course, stuff _can_ fail, but it works for the simple cases, and a
great deal of the not-so-simple cases. I did this for our cluster the other
day (130 wildly different databases, from 7.4 to 8.1) and it worked
flawlessly.

I do not really see why all the distributions could do something like this,
instead of mucking around with special statically compiled pg_dumps and the
like...

/* Steinar */
-- 
Homepage: http://www.sesse.net/