From: Joshua D. F. <jdf...@gm...> - 2006-02-14 17:45:29
|
On 2/14/06, Jeremy James wrote: > Joshua Daniel Franklin wrote: > > I unfortunately have a RAID 5 array on a 3ware 8506-4LP that > > lost a member, and now won't rebuild with a new drive (on Port #0). > > It gets about 60% done and reports a problem with one of the > > in-use drives (Port #2): > > > This is a common problem with RAID-5 - a failure on a second disk whilst > re-building the array. When we had this problem in the past, the best > solution is to dd the data from that disk to a new one (if dd breaks, > download dd_rescue [1] or similar instead which will not abort on bad > blocks). Swap that new disk into the array (you look like you're using > linux md, so as long as the superblock is copied it will be recognised > as the old disk - if you're using 3ware RAID this might be a lot > harder), rebuild the array, and run fsck on the filesystem when you're > done. Hopefully you'll only have lost a couple of sectors, so one or two > files. You might want to run fsck in a no-act mode first to get a > heads-up of what you might lose in the process. Thanks very much for the reply, it's what I was afraid of. We are using 3ware's hardware RAID, so I don't think I can fool it with the dd trick. Fortunately we've got good backups and the array is only about 1/3 full, so rebuilding isn't a big problem. I'm a little suprised that the 3ware controller can't try to skip bad sectors--I think I vaguely remember this feature in a Dell-branded aacraid controller. Or maybe that's a SCSI vs IDE thing. > Oh, and make sure you've got good backups... :) |