Message-ID: <4C946EE2.5020701@2ndquadrant.com>
Date: Sat, 18 Sep 2010 03:48:50 -0400
From: Greg Smith <greg@2ndquadrant.com>
User-Agent: Thunderbird 2.0.0.24 (X11/20100317)
MIME-Version: 1.0
To: Scott Marlowe <scott.marlowe@gmail.com>
CC: dbhandary@iii.com, pgsql-admin@postgresql.org
Subject: Re: files under pg_clog directories are missing
References: <4C940889.3000505@iii.com>	<4C9432DA.4000609@2ndquadrant.com>
	<AANLkTik2hQz6yx186=kz4=4u+ZgKHWmDdW974QnFFP2P@mail.gmail.com>
In-Reply-To: <AANLkTik2hQz6yx186=kz4=4u+ZgKHWmDdW974QnFFP2P@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Scott Marlowe wrote:
> As a followup to all this, the OP also needs to figure out why a
> crashed server could cause so much corruption.  This is usually the
> result of hardware lying about fsync status, i.e. consumer grade
> hardware pressed into service as a server.
>   

See http://www.postgresql.org/docs/9.0/static/wal-reliability.html for 
more information.  Nothing there has changed in 9.0 relative to the 
version being run here, but the docs for that version got a big upgrade 
over the earlier ones.

Bad memory is always a potential source for issues in this area too.  
Mostly if you don't know why the crash happened, either.  If you know 
for sure how it crashed, but only then found corruption, Scott's right 
that it's more likely to be a lying disk drive instead. 

-- 
Greg Smith, 2ndQuadrant US greg@2ndQuadrant.com Baltimore, MD
PostgreSQL Training, Services and Support  www.2ndQuadrant.us
Author, "PostgreSQL 9.0 High Performance"    Pre-ordering at:
https://www.packtpub.com/postgresql-9-0-high-performance/book