On Sat, 2007-01-27 at 06:38 +0900, John Summerfield wrote: > John R Pierce wrote: > > Mike Kercher wrote: > > > >> I got an email from smartd yesterday with the following error: > >> > >> SMART Health Status: SERVO IMPENDING FAILURE SEEK ERROR RATE TOO HIGH > >> [asc=5d,ascq=43] > >> > >> I understand this means the drive is probably going to fail. This > >> device is /dev/sda and is part of /dev/mapper/VolGroup00-LogVol00 > >> > >> Is it possible to replace the single failing drive and if so, what would > >> be the steps so that data isn't lost. > >> > > > > > > http://tldp.org/HOWTO/LVM-HOWTO/removeadisk.html > > > > /dev/sda, is that your boot device? that will greatly complicate > > things... and, your swap is probably not under LVM, /boot certainly > > isn't so there's moer on that drive than just part of VolGroup00 > > > > I think I'd opt for plan B... install an identical new drive on another > > SCSI unit, boot a standalone live CD, and DD image the whole raw > > physical drive /dev/sda, then swap this new drive to SCSI unit 0 and > > remove the failed one. > > > I discovered recently that getting exactly the same size (I am using > ATA) can be difficult, and two drives "the same size" can differ by a > few sectors. > > I'm sure bigger is okay, smaller is not (but you might be able to fiddle > with /boot and/or dispense with the swap partition. I prefer swap files > anyway)) Part of the reason I now leave the last couple percent un-allocated at the end of the drive. 80GB drives used to vary depending on the number of actual platters etc. I've done the CD boot and DD twice far successfully to a mirrored pair. IIRC if the geometry does not exactly match when you run fdisk on the drive it will ask to fix the problem. Paul