Message-ID: <4F9142C0.80008@gmail.com>
Date: Fri, 20 Apr 2012 13:04:32 +0200
From: Alexander Fortin <alexander.fortin@gmail.com>
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7;
	rv:11.0) Gecko/20120327 Thunderbird/11.0.1
MIME-Version: 1.0
To: pgsql-admin@postgresql.org
Subject: Re: pg_database_size() and actual disk space usage
References: <4F742D3A.8010501@gmail.com>
	<CA+h6AhgGgREbgpt2bobXUzunppWRbgii2TYMSksSbeYouM9Agg@mail.gmail.com>
	<4F743C47.1040907@gmail.com> <4F746282.3060406@innogames.de>
	<4F755EBE.2070908@gmail.com> <4F758796.6090000@innogames.de>
In-Reply-To: <4F758796.6090000@innogames.de>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

On 3/30/12 12:14 PM, Bernhard Schrader wrote:
> 
> Ok, well with Kernel 2.6.38 xfs introduced a new "feature" (i think it
> was dynamic preallocation) which caused some strange behaviour for me
> also, i discussed it on xfs mailinglist, maybe this gives you some hints
> of solving your different diskusages.
> 
> http://old.nabble.com/Problems-with-filesizes-on-different-Kernels-td33342207.html#a33342207
> 
> 
> At the end was the solution for me, but you can first check what i
> posted in the first message - testing base directory with du -sh and du
> --apparent-size -h
> 
> i hope this will help you.

Hi Bernhard,
just to say thank you to have pointed this out, it actually seems to be
related to this "dynamic speculative EOF preallocation" feature of XFS.

For the records, I've done a couple of checks and I can see clearly that
it's definitely unrelated to Postgres (apparently seems that I can
reproduce the change of data occupation on the XFS file system just
accessing those files).

Talk soon,

-- 
Alexander Fortin
http://about.me/alexanderfortin/