From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Hardy Subject: Re: RAID5 drive failure, please verify my commands Date: Sun, 16 Jan 2005 15:13:42 -0800 Message-ID: <41EAF526.2000707@h3c.com> References: <7C8198B3-67EA-11D9-9A87-00039363AEBE@bitart.com> <41EADBAE.6070400@h3c.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-raid-owner@vger.kernel.org To: linux-raid@vger.kernel.org List-Id: linux-raid.ids Gerd Knops wrote: > Seems to be a whole slew of bad sectors, here is what the log says: I typically get bad sectors in little mini-batches. Under a hundred - typically 40 or so in one chunk. There are 8 sectors per block, and I guess when a block goes, whatever took it out (some dust or something? a bad alignment of metal in the media?) gets the blocks near it too. Anyway, it still looks to me like its semi-normal. Definitely check out the bad block howto and (after removing the drive from the array) try doing a dd if=/dev/sda of=/dev/null bs=4096 skip=(whatever the sector is divided by 8) count=1 You should see the read errors in the log then, that verifies that you've got the right offset in the disk. At that point, do a dd if=/dev/zero of=/dev/sda bs=4096 skip=(sector address divided by 8) count=1 If you repeat the read after that block write, the read should succeed - you've told the drive you don't need that data any more (you wrote into the block) so it shouuld be able to reallocate If you take that single-block idea, and expand it to the first block address and a count= number that includes all the bad blocks, you should clear them all ?? Or the drive is shot :-). Without a SMART self-test and a check of the results where you can really interrogate the drive, its hard to say Glad this helps though -Mike