My terabyte raid1 —15 October 2008

My raid1 array with two 120G disks was getting filled up lately, so when usage reached 90% I decided it’s time to upgrade. To my great surprise prices have gone much lower than I expected: I got two 1T disks for the same price as the old 120G disks – $100 apiece.

It’s been a long time since I setup my raid1 array so I had to read up on man mdadm and mdadm --help.

1. Created a full backup of the old disks:
   cd /storage; tar cvpf disk200/raid1.tar raid1
2. Powered down the machine, replaced the disks, checked in BIOS they are recognized correctly.
3. Started up the machine, created a single primary partition on the two new disks with cfdisk, set the filesystem size to 1000G. This left a few hundred MB free space at the end of the disk, a safe thing to do in my experience.
4. Created the raid1 md device:
   mdadm --create /dev/md0 -l1 -n2 /dev/sdc1 /dev/sde1
5. The resyncing of two 1T disks takes time. An easy way to monitor the progress is with watch cat /proc/mdstat. It was more convenient for me to have a notification email sent to my mobile phone with while cat /proc/mdstat | grep resync; do sleep 60; done; cat /proc/mdstat | mail -s mdstat titan2x@mysupersecretmobile.com
6. Once the resync is done I created filesystem on it:
   mkfs.ext3 /dev/md0
7. In /etc/fstab filesystems are identified by UUID, not by traditional names like /dev/md0. To follow this new practice I updated the UUID in /etc/fstab for the partition according to the output of blkid.
8. mount -a complained about /dev/disk/by-uuid/... missing. This is fixed by restarting udev:
   /etc/init.d/udev restart
9. Mounted the partition with mount -a, and restored the files:
   cd /storage; tar xvpf disk200/raid1.tar