Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261734AbUKALNh (ORCPT ); Mon, 1 Nov 2004 06:13:37 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261747AbUKALNe (ORCPT ); Mon, 1 Nov 2004 06:13:34 -0500 Received: from 81-223-104-78.krugerstrasse.xdsl-line.inode.at ([81.223.104.78]:47507 "EHLO mail.sk-tech.net") by vger.kernel.org with ESMTP id S261734AbUKALN1 (ORCPT ); Mon, 1 Nov 2004 06:13:27 -0500 Date: Mon, 1 Nov 2004 12:13:51 +0100 (CET) From: Kianusch Sayah Karadji To: Neil Brown cc: linux-kernel@vger.kernel.org Subject: Re: Raid1 DM vs MD In-Reply-To: <16773.35825.194304.830001@cse.unsw.edu.au> Message-ID: References: <16773.35825.194304.830001@cse.unsw.edu.au> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2799 Lines: 62 Hi! >> Should I use MD or DM? > > I would say "md", but I am biased. :) > It is unlikely that support for useful functionality will go away. I > would prefer to see all arrays being assembled by an initrd-like thing > (initramfs??) but wouldn't dream of encouraging that until it was easy > to use and widely used. We aren't there yet. It seem that there has been some changes lately on the way when and by whom the RAID devices are created (especially in the 2.6.x stream) ... but one of the most frustrating experiences is the lack of decent documentation - or maybe some examples... Raid Support on DM seems not be documented at all ... and than there are some "SubTypes" like Raid Semi-Raid-Controllers (CMD649 etc.). Or maybe I'm searched the wrong way on google? Is there a central resource on the net for Linux & Raid I do not know of? >> I had one MD-Raid1 where a good copy of the mirror was overwritten by >> the bad (old) copy ... I lost 3 Month worth of data and I am expecting >> loosing a linux project and in the worst case - even a court case :( > > That sounds very unfortunate. Without knowing the details it is hard to > comment on why this might have happened and how it could have been > avoided, but with modern tools (mdadm) and a sufficiently modern kernel > (2.4 at least) this should never be able to happen (without deliberate > carelessness on the part of the sysadmin). Hmmm ... RedHat ES 2.x ... so no mdadm but raidtools ... The System was set up with HA-Linux so the sysadmin just had to initiate the TakeOver ... According to /dev/mdstat the System was okay ... but where do I see which copy is the "master" and where do I see in which direction the synchronisation is going to take place - before the raid is started? Also I did not find a way to start the raid with just one of the mirrors. As I said my biggest problem is the lack of documentation. >> Sooner or later I'l migrate from SW-Raid to a HW-Raid-Controller ... > > Many believe that that would not be a win. Personally I share that > view. HW-Raid-Controllers are not "open". md (and dm) SW-Raid is. Well ... yes ... but with HW-Raid users do not have to think very much. (I'm talking about REAL Raid Controllers - like 3ware, ICP, Adaptek - not about Soft-Raid-Controllers) Simply plug in the HardDisk und look how the system syncs the disc. And in my case I'm talking about mirroring 2TB Data ... MD takes up to 12Hours for syncing the raid - and stopping/starting MD during this time is always risky. Kianusch - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/