This week, when I noticed that 500GB drives had gone down to $99 at Canada Computers, I decided to upgrade my file server.

I use mdadm, a Linux-based software RAID solution, to handle the array. For the non-technical, this means that if I have four hard drives I can combine them into one big drive with a certain percentage of the space reserved for parity data. This parity data acts as a safety net: if any one of the hard drives die, the missing contents of that drive can be calculated from the data on the remaining drives. It’s magic!

When the two extra drives were installed, the array had to be grown to use them. The documentation for doing this is kind of scattered all over the place so I wanted to record the steps I went through in one place:

Notes

My raid device:
/dev/md0 (ext3 filesystem)
Existing drives in the array:
/dev/sdb, /dev/sdc, /dev/sdd, /dev/sde
Newly installed drives:
/dev/sdf, /dev/sdg

Add the two extra drives

mdadm –add /dev/md0 /dev/sdf
mdadm –add /dev/md0 /dev/sdg

Grow the array

mdadm –grow /dev/md0 –raid-devices=6

Increase the default speed of the growth (For me this meant the difference between taking 125 hours to complete and just 12 hours)

echo 1000000 > /proc/sys/dev/raid/speed_limit_max
echo 50000 > /proc/sys/dev/raid/speed_limit_min

Watch and wait for the growth of the array to finish

watch cat /proc/mdstat

Do a filesystem check

e2fsck -f /dev/md0

Resize the filesystem to match the new array size

resize2fs /dev/md0

Isn’t Linux ridiculously simple? If you’re using /etc/mdadm.conf don’t forget to update it with the new value for num-devices.