Two weeks ago, when nature was flexing its muscle in the northeastern USA in the form of hurricane Sandy, I was away from home on a business trip. When the alert email from my monitoring company arrived in my inbox informing me of the web server outage, I wasn't surprised. Power had been lost, the UPS had run out of juice and the server had gone silent.
But with the return of power, the server (which was setup to spring back to life) didn't come back and with me awaiting a return flight, the outage would go on a few more days. Diagnosing the server after returning home turned out to be a futile exercise. There were no hints as to why the server had failed to properly boot. So I just powered it up, repaired a few corrupt databases and thought that was the end of it.
But the server wasn't its old self, it kept on crashing with an ever-increasing frequency. Eventually I decided that the server had suffered critical, yet unidentified, damage to its hardware and last weekend I reluctantly replaced it with a newer box. Restoring a server is no walk in the park.
Over a year ago when I decided to host this site on my own server, I knew about the risks of self-hosting. The storm and the ensuing issues certainly proved some of those risks. There were loss of traffic, user inconvenience, and loss of Google ranking. A more robust setup might have averted some of that, but this website isn't quite at the point to justify that level of operation. Still I don't regret self-hosting. A hosted service can never match the same level of control and in the end it may not necessarily be that much more reliable either.