xfs
[Top] [All Lists]

Re: file corruption

To: stevew@xxxxxxxxxxxxxxx
Subject: Re: file corruption
From: "Stephen C. Tweedie" <sct@xxxxxxxxxx>
Date: 05 Apr 2004 12:20:52 +0100
Cc: linux-xfs@xxxxxxxxxxx, Stephen Tweedie <sct@xxxxxxxxxx>
In-reply-to: <200404050759.02927.stevew@xxxxxxxxxxxxxxx>
Organization:
References: <406AF7B6.6030405@xxxxxxxxxxxxxxx> <406CC518.1090204@xxxxxxxxxxxxxxx> <20040402015022.GA25936@xxxxxxxxxxxxxxxxxxxxxxx> <200404050759.02927.stevew@xxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
Hi,

On Sun, 2004-04-04 at 20:59, Steve Wray wrote:

> > some fs' will journal all data though (reiserfs and ext3 can do
> > this), but it often comes at a significant performance penalty for no
> > real gain (and sometimes causes other problems like seeing old/stale
> > data)
> 
> This is why I use ext3 with data=journal on /var/log

Why?  The default "data=ordered" mode for ext3 is 100% identical to
data=journal for the sorts of append-only file activity that you usually
see in /var/log.  The only visible difference between data=ordered and
data=journal that you can expect after a power-failure is that for
update-in-place operations, such as database updates, data=journal
preserves strict write ordering but data=ordered does not.  

For newly allocated data --- new files or appends to old files ---
data=ordered gives exactly the same benefit, without the performance
penalty of writing the data twice.

--Stephen



<Prev in Thread] Current Thread [Next in Thread>