Uptime is overrated
October 27th, 2009A long uptime is often spoken of as if it is a badge of honor, and many of us parade this number as if it means something. I’ll grant you that it feels cool to have a machine up for 383 days, but you give something up by allowing things to go so long. You lose the knowledge that your machine is going to boot in a state that you are happy with.
This frequent rebooting hypothesis has been kicking around in my head for a couple years, and tonight I finally had an experience that cemented it for me; I rebooted my server and one of my web sites was simply all messed up. The site was supposed to be a simple graphic design portfolio, but instead I was seeing network statistics generated by ntop; ntop?? What?? Why??
Here is what happened; this website is a Rails site using Mongrel running on the default port of 3000. In previous reboots there was no problem and the site came up fine. However, I installed ntop a few months ago and hadn’t rebooted since. Well, it turns out that by default ntop runs some kind of informational webserver on the same port. Since ntop is started before Mongrel the nice design site was replaced with network usage graphs.
Fortunately the timing of this reboot was at a time when it isn’t critical that the site is up, but had the machine had an unscheduled reboot during business hours I would have had to fumble around with this then.
Long story short; reboots aren’t so bad.
