Re: URGENT: Need help rebuilding iir RAID5 array with failed drive

On Sat, Jun 28, 2008 at 5:03 PM, Søren Schmidt <sos_at_freebsd.org> wrote:
>
> On 28Jun, 2008, at 23:10 , Garrett Cooper wrote:
>
>>
>> FWIW (I've discovered this through personal experience and reading a
>> lot of docs), the only way to get "hardware RAID" with iir and the
>> ICH9R chipset is through the Matrix Manager (either by creating one at
>> the BIOS level console or Windows -- bleh). FreeBSD spotted it as a
>> single drive (/dev/ar0), so at that point it was being managed by the
>> southbridge.
>
> OK, if you have and ICH9 you *do not and can not* use the IIR driver, the
> IIR driver is for HW based RAID cards from Intel (the devices show up as CAM
> disks). I wonder where you got the impression that the IIR driver was in use
> ? The ICHXX are all just plain (S)ATA devices, with the versions ending in
> 'R' having a different PCI ID so the BIOS can tell if you paid extra to get
> the software RAID "option" . Its all just marketing rubbish...
>
> This is the essential piece of info here.
>
>> Performance sucks, the array rebuild takes eons (16 hours for adding a
>> 1TB drive to an existing 4 x 750GB drive array with an Core 2 E6700
>> with 2GB RAM under Vista x64) and the rebuild console is _only_
>> available under Windows =(.
>>
>> At least it's keeping the filesystem intact though, long enough for me
>> to make a redundant copy of the data then move all this junk over to
>> another safe place while I grab DVD+R/W's and wait for my 3ware card
>> to come in the mail..
>
> I hate to bring this to you, but you will not have an intact filesystem when
> the rebuild is done, you have no parity data to rebuild from, just as the
> docs states.
>
> -Søren

Just as you predicted, the disk(s) are unreadable still, even after
the rebuild (and I hope to hell that it didn't destroy/disturb any
data on the surviving members).

I may just have to get it sent off for recovery (after I dd the
remaining members because there's still some usable data there --
bloody 2.7TB ... grr) -- apart from that I don't think I have a
choice.

Oh, where's this "document" of which you speak all the time? I don't
see anything about parity support not being available in either the
geom or iir manpage, or the iir sourcecode. If someone needs a helping
hand inserting that logic into the kernel, I'll gladly assist so
others don't follow my footsteps and fsck up their data.

For once I wish I had been running Linux or Windows (at least) :(...

Thanks,
-Garrett