From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Fri, 24 Aug 2001 19:44:56 -0700 From: David Rees Subject: Re: [linux-lvm] LVM, rawio, ext2? Message-ID: <20010824194456.A25251@greenhydrant.com> References: <20010823014012.A32700@greenhydrant.com> <46CEA578.B03DDD14@qis-systemhaus.de> Mime-Version: 1.0 Content-Disposition: inline In-Reply-To: <46CEA578.B03DDD14@qis-systemhaus.de>; from heinrichs@qis-systemhaus.de on Fri, Aug 24, 2007 at 11:31:36AM +0200 Sender: linux-lvm-admin@sistina.com Errors-To: linux-lvm-admin@sistina.com Reply-To: linux-lvm@sistina.com List-Help: List-Post: List-Subscribe: , List-Unsubscribe: , List-Archive: List-Id: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-lvm@sistina.com On Fri, Aug 24, 2007 at 11:31:36AM +0200, Dirk Heinrichs wrote: > David Rees wrote: > > > > Hi, > > > > I'd like to create an ext2 filesystem over a raw io device, is this even > > possible? (If not, skip the rest of this message! ;-) > > Why do you want to do that? > > These are two opposite things. One uses a raw device where an > application has it's own storage layout, like databases. > > If you just want to make shure that your files on an ext2 are written > immediately, then mount with sync option. Well, the version of Sybase I'm running (11.9.2) does not support Linux raw partitions. Sybase 12.5 reportedly does, but I have a number of servers running 11.9.2 and don't want to upgrade yet. Sybase 11.9.2 only supports database data put onto filesystems on Linux, so I was hoping to create a situation where writes to the filesystem would not be buffered by the operating system since Sybase has it's own buffers. Buffering IO to the filesystem also has the nasty side-effect of putting increased VM pressure on the system and with 2.4.X kernels this leads to poor performance. (See the threads on linux-kernel describing huge slowdowns with Oracle and 2.4.x when moving from 2.2.x). Until I can validate my applications on Sybase 12.5 using rawio, I was looking for a way to prevent filesystem IO from being buffered. If anyone has any ideas, let me know! Thanks, Dave