DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=mime-version:in-reply-to:references:date:message-id:subject:from:to
	:cc:content-type;
	b=mOiyxaQyiv1QmI8pSPGBOf0Q6M7OaapntoWkAbb36k9BwGz1Rj9HvgL4bKX1foenjd
	dAlQvp1Ch3t2hlvAOq2PouWBRphelnMOItuWXysZnJrLrJ5Icak5hKm1fU/QCiETlMHI
	VH9RhZGTW5EeRHEZ2yx4rS6dIUS4VIYoF7zRk=
MIME-Version: 1.0
In-Reply-To: <op.u944juqoeorkce@localhost>
References: <4BA23936.2030507@gmail.com> <4BA596D2.4050406@squeakycode.net>
	<20100323185232.GD6662@rice.edu> <5342.1269372121@sss.pgh.pa.us>
	<4BA934B9.9070401@2ndquadrant.com> <4BA9D39B.6020103@gmail.com>
	<1269421198.11473.50.camel@hvost>
	<4BABC1E0.7010607@mansionfamily.plus.com>
	<op.u944juqoeorkce@localhost>
Date: Thu, 25 Mar 2010 14:38:07 -0600
Message-ID: <dcc563d11003251338m1369b4efh2efa600c9c8fe7dd@mail.gmail.com>
Subject: Re: mysql to postgresql, performance questions
From: Scott Marlowe <scott.marlowe@gmail.com>
To: Pierre C <lists@peufeu.com>
Cc: James Mansion <james@mansionfamily.plus.com>,
	Hannu Krosing <hannu@2ndquadrant.com>,
	Yeb Havinga <yebhavinga@gmail.com>, Greg Smith <greg@2ndquadrant.com>,
	Tom Lane <tgl@sss.pgh.pa.us>, "Ross J. Reedstrom" <reedstrm@rice.edu>,
	pgsql-performance@postgresql.org
Content-Type: text/plain; charset=ISO-8859-1

On Thu, Mar 25, 2010 at 2:29 PM, Pierre C <lists@peufeu.com> wrote:
>> Hannu Krosing wrote:
>>>
>>> Pulling the plug should not corrupt a postgreSQL database, unless it was
>>> using disks which lie about write caching.
>>>
>> Didn't we recently put the old wife's 'the disks lied' tale to bed in
>> favour of actually admiting that some well known filesystems and saftware
>> raid systems have had trouble with their write barriers?
>
> I put a cheap UPS on the home server (which uses Software RAID) precisely
> because I don't really trust that stuff, and there is also the RAID5 write
> hole... and maybe the RAID1 write hole too... and installing a UPS takes
> less time that actually figuring out if the system is power-loss-safe.

Very true, a UPS might not cover every possible failure mode, but it
sure takes care of an aweful lot of the common ones.