- Apache server wouldn’t restart on reboot. Answers I was look for were “look in system logs for errors,” “try and restart it manually,” and where those logs and controls might be. The kernel patch was a red herring, it wasn’t even related. Extra points for those who tried to fix it so it WOULD restart next time the box was booted. This comes from several personal experiences where the /etc/init.d/ script for Apache was not executable because the packages we pushed out forgot that step.
- The primary database server. This was a security question. I am looking for, “How does he know what the system logs say if you’re the only one with root access?” I left several hints that this was an attempted fraud/hack/social engineer attempt by placing undue pressure on the applicant. I would have accepted any answer where the admin checked the primary server first, because he would have found nothing wrong with it. I’d even take looking at the backup to see what was wrong with it before I failed it over. Years ago, someone hacked the backup DNS server in our UK offices, but not the primary (which had been properly patched). No one tried to get us to fail over, I just added that part myself for the question.
- The upgrade database package. I was just asked this in an interview, and my guess was right: someone pushed out a packages that wrote to SCSI drives, and two machines had ATA drives in them. I would have accepted any answer that looked at the /etc/fstab.
- The forums keep crashing. This happened to one of my own projects. The swap file was not mounted (it was improperly spelled in the fstab), so when a runaway PostGreSQL fulled up the RAM, after 4GB, it had no virtual memory to write to, and the whole box hung. I would have accepted any answer that used TOP, ps, looked at the fstab, system logs, etc...
- Boxes that can’t ssh from A to B. An “in.sshd” was not properly added to /etc/hosts.allow on B. I would have accepted any answer that mentioned that, or even tried iptables, just to recognize that it wasn’t a hardware network issue. This happened to me when a package with a “generic hosts.allow” had a default deny policy, and suddenly, this box couldn’t be reached by anything put pings and ssh from an admin host.
- Linux box shares. I was looking for help desk skills, like, “What share are you trying to reach, what login do you have, and what is the error?” Maybe that box isn’t really Linux after all. This happened to me a lot. Just because I am “the Linux guy” doesn’t mean I control all Linux boxes everywhere. I would have also accepted samba troubleshooting skills.
My guess is, based on lack of serious responses, these were too hard and unfair. Is this true? Please be honest.