From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bill Davidsen Subject: [RFC] Additional info on mdadm 'check' Date: Thu, 16 Aug 2007 08:21:18 -0400 Message-ID: <46C4413E.3060302@tmr.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Sender: linux-raid-owner@vger.kernel.org To: Linux RAID List-Id: linux-raid.ids I was wondering if there would be enough value in saving more information out of 'check' to justify the effort of coding it. Particularly in the area of faster repair. If the check operation optionally left a bitmap of stripes which were bad, if the admin decided that a repair was appropriate, then only the stripes identified by check would need to be processed. Because of possible additional errors, etc, that should be optional as well. This opens the door to more intensive repair operations, such as voting on more-than-two-way mirrors, and extensive test and try recovery on RAID-6, where no read error is found but the data doesn't match. By assuming chunks are bad, one at a time, a correct sector might be identified and written back. Note: this is just an idea at the moment, I'm just doing a feasibility study here. ;-) -- bill davidsen CTO TMR Associates, Inc Doing interesting things with small computers since 1979