Usually when creating a RAID array, it is expected to transfer the data outside before creating the RAID array because creating a new RAID array is a destructive process. However, that does not necessarily apply to a RAID 1 (mirror) array that can be created using a drive with already existing data given that all the requirements for a RAID 1 array are satisfied.

A side-effect advantage of this procedure is that the data has to be copied only once after which the data can be used whilst the RAID array synchronises.

• Two hard drives of the exact same size.
• mdadm (install with aptitude or apt-get under Debian-like system).

The tutorial is written under the following assumptions:

• The hard-drive that already contains the data will be referred to as /dev/sda.
• The spare drive that will become part of the array will be referred to as /dev/sdb

In order to copy over the existing partition table from the drive already containing the data (/dev/sda) to the spare drive (/dev/sdb) we issue:

sfdisk -d /dev/sda | sfdisk /dev/sdb

where:

• -d tells sfdisk to dump the partition table
• /dev/sda is the drive containing the data
• /dev/sdb is the spare drive

Next we create a RAID 1 array with two drives from the spare drive that does not contain any data:

mdadm --create /dev/md0 --level=mirror --raid-devices 2 /dev/sdb1 missing

where:

• /dev/md0 is the RAID device to use
• --level=mirror is equivalent to RAID 1 mirror
• --raid-devices 2 specifies that the array will contain two devices
• /dev/sdb1 is the first partition of the spare drive that does not contain any data
• missing tells mdadm to create the RAID 1 array with the second drive missing

Now we would have created a RAID 1 array referenced by the /dev/md0 device with only the spare drive /dev/sdb being a member. To confirm, issue:

cat /proc/mdstat

which should output something like:

md0 : active raid1 sdb1[0]

The next step is to format this RAID device with only one member. For instance, to use the Ext4 filesystem, we would issue:

mkfs.ext4 /dev/md0

Finally, what we are going to do is copy over the existing data from /dev/sda1 into the newly created RAID array with one missing drive. In order to do that, we mount /dev/md0 to a mount-point:

mount /dev/md0 /mnt/raid1

where:

• /dev/md0 is the RAID array we just created
• /mnt/raid1 is a folder - can be any empty not-used folder

Then, we mount the drive that already contains data (/dev/sda1) to some other mount point:

mount /dev/sda1 /mnt/trans

And copy the data over from the drive already containing the data /dev/sda1 into the RAID array:

rsync -P -av /mnt/trans/ /mnt/raid1

which, depending on your situation, could take a while.

After all the data has been copied over, unmount all the mount points (/mnt/raid1 and /mnt/trans in this tutorial):

umount /mnt/raid1
umount /mnt/trans

and then add the drive from which we copied the data over (/dev/sda1) to the RAID 1 array:

mdadm --add /dev/md0 /dev/sda1 

In order to watch the RAID array synchronising both drives you can issue:

cat /proc/mdstat

as well as get other details about the RAID array with:

mdadm --detail /dev/md0