Data redundancy (RAID mirror)

Today diglloyd.com was down briefly while the server was moved. When it rebooted, one of the hard drives did not spin up, simulating a disk failure. Fortunately, the hard disk hadn’t actually failed, rather it was a cabling issue. And even better, the diglloyd.com web server runs on a 2-drive mirrored volume so that failure of one hard disk won’t take the site down.

A mirrored volume consists of two or more hard disk partitions (slices, eg portions of a hard disk), each residing on different physical hard disks. The same data is written redundantly to all the member partitions of the mirrored volume (“carbon copies”). When a hard disk fails, all remaining partitions continue operating as before; there is no loss of data. When the bad disk is replaced or brought online again, it is “rebuilt” using the data on the other mirror partitions.

Mac OS X has built-in support for mirrored volumes. Seen below is the progress shown while Disk Utility rebuilds the errant disk partition. (“Degraded” means that one or more disks have failed). For even more protection against disk failures, a mirrored volume could use more than the minimum of two disks, but barring an Act of Dog that might ruin all the disks at once, odds of two hard drives failing within a few days of each other is rather small, so two disks should suffice for most uses.

Click to see full image

I’ll have much more to say on the subject of storage in the future. But in particular, Mac Pro users with empty drive bays might find that a mirrored volume offers simple yet highly effective protection against disk failure. Setting up a mirrored volume is easy, and one can even boot from it, offering protection for both the system and its data.