On 3/6/07, Leen de Braal <[EMAIL PROTECTED]> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> The Tuesday 2007-03-06 at 13:28 +0100, Leen de Braal wrote:
>
>> Still asking myself how this could have happened? Any idea?
>
> Look at the logs... it's the only way. It could be a glitch. There is a
> temporary problem sometime, a disk is removed, and it awaits manual
> intervention. It will automatically activate an spare if available,
> though.
>

Found:

Mar  5 00:17:14 linux kernel: hda: dma_intr: status=0x51 { DriveReady
SeekComplete Error }
Mar  5 00:17:14 linux kernel: hda: dma_intr: error=0x40 {
UncorrectableError }, LBAsect=273480054, high=16, low=5044598,
sector=273480053
Mar  5 00:17:14 linux kernel: ide: failed opcode was: unknown
Mar  5 00:17:14 linux kernel: end_request: I/O error, dev hda, sector
273480053
Mar  5 00:17:14 linux kernel: raid1: Disk failure on hda3, disabling device.
Mar  5 00:17:14 linux kernel:   Operation continuing on 1 devices
Mar  5 00:17:14 linux kernel: raid1: hda3: rescheduling sector 271343408
Mar  5 00:17:14 linux kernel: RAID1 conf printout:
Mar  5 00:17:14 linux kernel:  --- wd:1 rd:2
Mar  5 00:17:14 linux kernel:  disk 0, wo:1, o:0, dev:hda3
Mar  5 00:17:14 linux kernel:  disk 1, wo:0, o:1, dev:hdb3
Mar  5 00:17:14 linux kernel: RAID1 conf printout:
Mar  5 00:17:14 linux kernel:  --- wd:1 rd:2
Mar  5 00:17:14 linux kernel:  disk 1, wo:0, o:1, dev:hdb3
Mar  5 00:17:14 linux kernel: raid1: hdb3: redirecting sector 271343408 to
another mirror

Is the above telling me that hda3 was removed from the mirror because
of a single bad sector?

That seems extremely aggressive.

I know there is some LKML discussion of needing to have MD
automatically detect the above and simply rewrite the failed sector
with data from the good mirrored sector.

During the write /dev/hda should re-map the failed sector and continue
running fine.  (ie. All disk sector remapping for failures happens on
writes AIUI.)

If a disk is failed after a single sector read error currently I can
see why the kernel developers are looking into alternate ways to
handle the situation.

Greg
--
Greg Freemyer
The Norcross Group
Forensics for the 21st Century
--
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to