From mboxrd@z Thu Jan 1 00:00:00 1970 From: Phil Turmel Subject: Re: Specifying a data offset? Date: Wed, 08 May 2013 16:44:20 -0400 Message-ID: <518AB924.2010002@turmel.org> References: <1368044274.3368.364.camel@148> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1368044274.3368.364.camel@148> Sender: linux-raid-owner@vger.kernel.org To: tbayly@bluehost.com Cc: linux-raid List-Id: linux-raid.ids On 05/08/2013 04:17 PM, Tregaron Bayly wrote: > We have lots of mirrors all over our network and over time this means > arrays created by different versions of mdadm. In turn this means that > we have all sorts of data offsets, including a 2.3.5 mdadm version that > offset the data 262,144 sectors. I can't fathom why I need my data to > start 128M from the beginning of the disk, but whatever. > > Inevitably an admin decided that doing a --create over the top of an > existing array was a good and necessary thing and when finished we > couldn't find the filesystem. Bringing it back online was simply a case > of finding the correct offset, locating an mdadm version that would > specify it properly and doing another create. Through the whole > exercise I found myself wondering why there isn't a command line > argument to specify the data offset if something other than the default > is desired. I think that implementing the argument and reading the > value in init_super wouldn't be terribly difficult so I must be missing > an obvious reason. Can someone explain to me why this feature isn't > desirable? This desirable and has been implemented. Check out the development version of mdadm and look at the "--data-offset" option. It does not exist in a released version yet, but it has been cooking since October. Phil