Re: Borked MD RAID...

[Jeffrey Ross <jeff bubble org>]

Lonni J Friedman wrote:
> Is it always the same disk that gets marked offline ?  Perhaps the
> disk is actually bad?
>
> On Sun, Nov 30, 2008 at 4:20 PM, Eitan Tsur <eitan tsur gmail com> wrote:
>   
> > Even if I remove that it still happens.
> >
> > On Sun, Nov 30, 2008 at 3:53 PM, Lonni J Friedman <netllama gmail com>
> > wrote:
> >     
> > > If you only have 3 disks, then you can't have:
> > > spares=1
> > >
> > >
> > >
> > > On Sun, Nov 30, 2008 at 3:11 PM, Eitan Tsur <eitan tsur gmail com> wrote:
> > >       
> > > > > # mdadm.conf written out by anaconda
> > > > > DEVICE partitions
> > > > > MAILADDR root localhost
> > > > >
> > > > > ARRAY /dev/md0 level=raid5 num-devices=3 spares=1
> > > > > UUID=0c21bf19:83747f05:70a4872d:90643876
> > > > >           
> > > > If I switch the "DEVICE partitions" with "DEVICE /dev/sdb1 /dev/sdc1
> > > > /dev/sdd1", drives no longer are allocated as spares, however the array
> > > > still seems to rebuild every boot.
> > > >
> > > > I don't remember the specifics of what was in /proc/mdstat at the time,
> > > > but
> > > > currently the array is being rebuilt.  I'll reboot after it is complete
> > > > to
> > > > give you a copy of it.  Basically it allocated the dropped drive as a
> > > > spare
> > > > which I'd have to mdadm -stop and mdadm -add to the original array after
> > > > every boot, manually.  Give me an hour or two and I'll get you the
> > > > output of
> > > > mdstat.
> > > >
> > > > On Sun, Nov 30, 2008 at 2:41 PM, Lonni J Friedman <netllama gmail com>
> > > > wrote:
> > > >         
> > > > > On Sun, Nov 30, 2008 at 2:38 PM, Eitan Tsur <eitan tsur gmail com>
> > > > > wrote:
> > > > >           
> > > > > > I just recently installed a 3-disk RAID5 array in a server of mine,
> > > > > > running
> > > > > > FC9. Upon reboot, one of the drives drops out, and is allocated as a
> > > > > > spare.
> > > > > > I suspect there is some sort of issue where DBUS re-arranges the
> > > > > > drive-to-device maps between boots, but I am not sure... Just kind of
> > > > > > annoying to have to stop and re-add a drive every boot, and wait the
> > > > > > couple
> > > > > > hours for the array to rebuild the 3rd disk. Any thoughts? Anyone
> > > > > > else
> > > > > > encountered such an issue before? What should I be looking for? I'm
> > > > > > new
> > > > > > to
> > > > > > the world of RAID, so any information you can give may be helpful.
> > > > > >             
> > > > > What's in /etc/mdadm.conf, /proc/mdstat and dmesg when this fails ?
> > > > >           
>
>   

check and make sure that the UUID number your specifying in your 
/etc/mdadm.conf file is correct.  You can verify the UUID numbers by 
typing "ls -l /dev/disk/by-uuid"

Next verify that the UUID numbers in your /etc/mdad.conf file stored in 
the initrd file is correct, you'll have to extract the initrd file
with cpio.  I don't remember the full procedure but you should be able 
to find it pretty easily with your favorite search engine.

Jeff