Nicollet.Net

Today, I ran rm -rf *. As root. On a production server.

The problem is not that I lost any important data—my surgical strike removed precisely the outdated files that had to be erased, and nothing else.

The problem is that I did it unconsciously. I did not stop to check that I was doing the right thing, which means that had my pwd been off by an inode or two I would have blasted important files away instead of junk. And would have spent the rest of the evening and night restoring them. And when I noticed this, moments after doing it, I went into a short-lived panic as I checked everything after the deed. Those were my two seconds of fear for the week, I guess.

Being a system administrator is all about being permanently afraid of the next thing that will happen. When you don’t mistype rm -rf *~ as rm -rf * ~ or overwrite original data with an incorrect tab-completed pipe-to-file, you end up with security holes that can and will be exploited at one point.

Did you know that changing an user password in the mysql.user table only takes effect on the next MySQL reboot? And that if the password is not a valid hash, MySQL assumes no password is required? Not all security holes appear right away.