[TriLUG] Problem with LVM

Rick DeNatale rick.denatale at gmail.com
Tue May 2 16:06:44 EDT 2006


On 5/2/06, T. Bryan <tbryan at python.net> wrote:

>
> > 1) Partition /dev/sdb like /dev/sda
> >
> > 2) Create the raid array with a missing drive
> > $mdadm --create /dev/md0 --level 1 --raid-devices=2 missing /dev/sdb5
>
> Yeah.  I saw these kinds of instructions for root filesystems.  Since I wasn't
> doing a root filesystem, I thought that I could just create the RAID with
> both disks in the mirror and then work on /dev/md0.  Like I said, I actually
> used /dev/localvg/lv_local1 for a few weeks before doing a dist-upgrade and
> upgrading my kernel.  I didn't seem to have any problems until then.

I don't think that this has anything to do with whether the raid array
is root o something elses.

I'm no raid/lvm guru, but I wouldn't expect md to be able to join two
supposedly "identical" disks into a raid 1 configuration. Note the
last question and answer in
http://linas.org/linux/Software-RAID/Software-RAID-3.html  The md
drivers actually muck with the disk formatting and use some space from
the partitions for the superblock.  Besides that, even though the two
filesystems might have identical contents when looked at through the
filesystem, md makes them identical at a lower block-oriented level. 
MD makes the partitions comprising the raid array look like a single
block device to everything above.  I think that this means that you
need to make the filesystem on the md device, not on the physical
devices.

The reason I created the raid array with a missing drive was because
one of the drives I was using for the raid array already had the data
I wanted in the raid array.  This strategy let me make the raid array,
copy the data to it, and then add the now available drive to the
array.

--
Rick DeNatale

Visit the Project Mercury Wiki Site
http://www.mercuryspacecraft.com/



More information about the TriLUG mailing list