To: Matthew Wakeling <matthew@flymine.org>
cc: pgsql-performance@postgresql.org
Subject: Re: I/O on select count(*) 
In-reply-to: <Pine.LNX.4.64.0805151421190.16756@aragorn.flymine.org> 
References: <920841.71690.qm@web44808.mail.sp1.yahoo.com>
	<482B1D5F.EE98.0025.0@wicourts.gov>
	<Pine.GSO.4.64.0805142133250.27290@westnet.com>
	<20080515020553.GN9838@alvh.no-ip.org>
	<Pine.LNX.4.64.0805151245360.16756@aragorn.flymine.org>
	<482C320F.9070908@enterprisedb.com>
	<Pine.LNX.4.64.0805151421190.16756@aragorn.flymine.org>
Comments: In-reply-to Matthew Wakeling <matthew@flymine.org>
	message dated "Thu, 15 May 2008 14:38:48 +0100"
Date: Thu, 15 May 2008 10:52:58 -0400
Message-ID: <4758.1210863178@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Matthew Wakeling <matthew@flymine.org> writes:
> Hmm. That problem is what WAL full-page-writes is meant to handle, isn't 
> it? So basically, if you're telling people that WAL full-page-writes is 
> safer than partial WAL, because it avoids updating pages in-place, then 
> you shouldn't be updating pages in-place for the hint bits either. You 
> can't win!

This argument ignores the nature of the data change.  With a hint-bit
update, no data is being shuffled around, so there is no danger from a
partial page write.

A disk that leaves an individual sector corrupt would be a problem,
but I don't think that's a huge risk.  Keep in mind that disks aren't
designed to just stop dead when power dies --- they are made to be able
to park their heads before the juice is entirely gone.  I think it's
reasonable to assume they'll finish writing the sector in progress
before they start parking.

			regards, tom lane