Friday, January 2, 2009

Trust No One

One of the TV shows I like the most is House. As many people point out, each episode is pretty much the same and yet it's one of very few shows that improve as seasons go on. Anyways, one motif that appears in almost every episode is that people shouldn't be trusted under any circumstances, every patient House treats has some terrible secret lurking in the dark.

Users are sometimes like patients, they don't necessarily have a dark secret (although they can occasionally forget to tell you an important detail) but they definitely cannot be trusted. That's why the very first step of handling a problem is watching it reproduce with your own eyes, there are a couple of reasons I can think of for doing so: first there's the Sysadmin Effect - never underestimate it. And then there's the fact that the reported problem is often not the issue at all (or not at all an issue). Examples I've heard of and experienced myself range from "The system is down" when working on a computer disconnected from electricity to calling hysterically about "a critical application is down" when the application is neither important, nor down. Obviously there are more examples in the mid-range but that's just a blog post not a thick book.
I have a special place in my heart for the guys that have some good advice, like the totally non-technical guy that advised me to add some memory to my servers to improve performance or the one that had an idea about how to fix the application (the OID) when the issue was with his OS.

So remember, always begin with the first step.