Backups: In Praise of Paranoia

Recently I had a disk drive failure, but I didn’t lose any data at all. In part that’s because that disk was “mirrored” to another disk using “RAID 1” or “mirroring”. In mirroring, software or hardware keeps two identical copies of a disk up-to-date as you go about your work, if one goes, you can replace the failed disk and (within a few hours) you can rebuild the mirror. It’s a good technology,and it allows you to continue to work when failure happens. Even so, I moved quickly to replace the disk and restore the mirror, despite having additional backups.

Now, a couple years back a few friends and I spent a few weeks hiking and photographing in Patagonia, often driving on rarely-used, rocky dirt roads. We were only a little surprised then, coming into El Calafante that we’d had a flat, one we fortuitously noticed just before we descended a step, winding cliff road. We easily replaced the tire and then immediately went and got the flat tire repaired in town, we’d never have considered returning to the wild without having a working spare. But that same advice applies equally well to failed backups.

That night in El Calafante, I had another opportunity to listen to my paranoia—I woke up about 2am to hear the sound of sizzling–the sizzling of the power brick on my laptop. Now, when I travel I usually have two separate ways of backing up images from a trip, usually one copy goes onto the laptop, and the other usually ends up on a portable storage device such as the excellent ::amazon(“B001CSPL7Q”,”Sanho Hyperdrive”):: units, or in this case a similar unit from another manufacturer. The loss of the laptop power supply thus presented me with a problem, I knew the images I had were okay (the laptop still worked), but I guessed I didn’t have enough power to use it for the rest of the trip. I immediately asked myself, “how can I get a second backup of any new images?” It wasn’t hard to answer that question, we were about two-thirds of the way through the trip and I realized I had enough memory cards that I could use the cards themselves as secondary backup for the remainder of the trip. I’m glad I did, too—upon returning to the US I found that the storage gadget had a few disk failures, because of my paranoia, I didn’t lose that last third of my Patagonia shoot.

This sort of “fix failures now” reflex is doubly warranted with RAID mirrors. Because such mirrors are usually constructed with nearly identical drives that get identical usage in a nearly identical physical environment, once one disk in a mirror fails it’s not uncommon for the second to fail soon thereafter. RAID is not a panacea. I could have lost the second disk as I was reconstructing the new one, and if that’d happened I would have had to go to offline backups (which would have been fine), but getting the RAID mirror restored was a lot less work. As if to prove the value of my paranoia, a couple weeks later that second original drive did fail, too, and I quickly replaced and mirrored it, I’m now running on a pair of fairly new disks.

It’s easy to remember that backups are there to help save you from failures. Just remember that backups can fail too, and having backup plans for your backup plans, knowing what you will do when something inevitably does fail, is essential.