A whole pile of preposterous questions in this list, except for this one :
“Emacs or vi?”
Why wouldn’t that be a good question? It should invoke an emotional response and the answer for most environments I have been in is:
Emacs or vi? vi of course, because it is on nearly every system by default, where as emacs is not. You can not, most times, go around installing your favorite editor on every system you have to admin.
If you need emacs, or some other editor/ide/whatever for large coding projects which you do on your local workstation, notebook, or management server… fill your boots.
If you need to make a quick change to /etc/fstab on a random server in the datacentre, you had damned well better know how to drive vi adequately to do that.
Knowledge of ssh, screen, and rsync are also absolute requirements, as also is the ability to “think right”.
Looks like I will have to think up one of my past ‘interesting’ problems that I have had to solve and use that as a way of probing a candidates thinking abilities tomorrow, when they show up for their 2nd interview (I wasn’t in the first one). Something like this problem someone had with Macs and Lotus Notes.
Sometimes the best questions are the simplest ones. I came to a revelation when you and I were last working together, it was a sad revelation, and it consisted of two statements:
1. Common Sense… ISN’T
2. People are stupid.
This has been confirmed in a manner of sorts by this posting on Coding Horror: http://www.codinghorror.com/blog/archives/000781.html
So, I will suggest a simple question whose answer should provide some insight into the level of understanding and comprehension of what is going on under the covers. I know you are familiar with this one… and all the hidden “gotchas” related to it.
You are working with a team of developers that has chosen to use a directory and a “file watcher” to facilitate data sharing mechanism. In the simplest form, one application writes a file, and other reads it and then removes it once it has processed it. What issues are there with this scheme? What are the solutions? What are the alternatives?
What I like about this question is that it can show how much a sysadmin understands about concurrency/threads, the impact of polling, file system caches, error detection, failure recovery, and many others.
The solutions aren’t difficult, they are pretty simple, but the why of the solution (as in why do you do it that way) can be quite telling.
camz
March 22nd, 2007
We didn’t get to the point where I could pose a question or scenario like that. The candidate volunteered plenty of fodder for discussing their work methods and thinking patterns.
The scenario you mention, the one we have worked through together in the past, is a good one to bring up next time though. I’ll keep it in mind, thanks.
I was also thinking of the “whats one of the worst situations you have been in, how did it happen, and how did you recover from it” question.
I’ve got a couple to throw out as starting points for discussion but I think the best one (at the moment, I may reconsider) was when we were moving to a new office location and had a new server to get up and running. I spent about three 18 hour days in a row getting that machine up and running with Solaris, Oracle, WebLogic, etc. It was to be the new production server that we would dump/load the data to and start using as soon as the SDSL was in place, so as to provide a smooth cutover.
I was tired. I had just proposed that am. My mind wasn’t in the right place.
Second last item on the server prep list … oh yeah, I forgot to mirror the drives. How does Solaris Disksuite do this again?? Zap. Blank mirrored overtop of the working drive. Whaaaa!
Last item on the todo list was, of course, backup system.
Once burned, twice shy, I always try to mirror drives first before any further work is done these days. Of course I absolutely am mind boggled as to why Sun does not have hardware mirroring by default in even their lowest end 1U machines. HP does (story to follow relating to upgrading raid-1 drives on HP DL360 from 72GB to 300GB with minimal downtime).
trever
March 23rd, 2007