You may have noticed that my website has been pretty barren lately. I suffered massive data loss when the hard drive in my home server died. I had been running the server as a virtual machine in the hopes that as long as I backed up the disk image I would be safe from hardware failures. I would have been correct if I had bothered to back up the disk image. So in late January of this year the drive failed, the server crashed and I lost everything.
When I realized the server wasn’t coming back on-line, I put a simple page up so that I wouldn’t lose my place at the top of Google for “Chris Moylan” searches. I also faced a tricky recovery process. If it had just been my data I would have shrugged it off and started fresh. But I was also hosting sites for friends, including the blog of my friend Matt who had years of entries and images.
I managed to get everything back in some form or another. Luckily I recovered all of Matt’s data and returned it to him. He promptly set up his blog on a different host and I don’t blame him at all. I decided to pay for email hosting (I had been hosting it myself) and found an awesome provider. In the future, if I want to host my own email server, I can use their service as a fallback. I took the time to migrate my site over to Hyde, a static site generator. I had been using Pylons, a Python MVC framework, but decided I didn’t need all of that complexity.
I had some free time last weekend so I set the server up again. I’m going with the same setup - a virtual server running on top of a physical server. Unlike last time, I am going to stick with some sort of backup schedule or add a RAID setup. This incident hasn’t deterred me from running a server out of my closet. It was a great learning experience, one that I plan to continue. But I now realize that any system with non-trivial uptime has a non-trivial risk of catastrophic failure. If you acknowledge this and plan for the day when your data goes out the window, it’s not big deal.