Date: Sun, 4 Apr 2004 13:53:13 -0700
From: Chris Wedgwood <cw@f00f.org>
To: Steve Wray <stevew@catalyst.net.nz>
Cc: linux-xfs@oss.sgi.com
Subject: Re: file corruption
Message-ID: <20040404205313.GA20703@dingdong.cryptoapps.com>
References: <406AF7B6.6030405@dniq-online.com> <406CC518.1090204@dniq-online.com> <20040402015022.GA25936@dingdong.cryptoapps.com> <200404050759.02927.stevew@catalyst.net.nz>
Mime-Version: 1.0
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <200404050759.02927.stevew@catalyst.net.nz>
Content-Transfer-Encoding: 8bit
Sender: linux-xfs-bounce@oss.sgi.com
Errors-to: linux-xfs-bounce@oss.sgi.com
Precedence: bulk
Content-Length: 856
Lines: 24

On Mon, Apr 05, 2004 at 07:59:02AM +1200, Steve Wray wrote:

> This is why I use ext3 with data=journal on /var/log

By default syslog writes synchronously...  however many configurations
have some files marked to write asynchronously.  Does "data=journal"
buy anything over synchronous writes?

> The advantage is that in event of a kernel panic or other hard
> lockup, one can actually find some useful hints in the logs as to
> what went wrong, instead of 'garbage binary data'.

Synchronous writes are probably more useful still.

> I wish XFS had an *option* to journal data...

There is already a massive amount of data written to the log (profile
how much IO there is to a log during a large "rm -rf" or something),
making this worse for a small number of corner-cases where alternative
solutions exist doesn't seem entirely useful as-is.


  --cw