Server Madness

I am a FreeBSD god. I can administrate my way in and out and upside down a FreeBSD box like mad, and do good work, too. Computers used to frustrate me endlessly until one day in 2000 I installed FreeBSD 4.0 on an unwanted machine and all my problems magically disappeared. It just made sense. And it worked. And the documentation was so complete that I hardly ever had to search the web for how to do stuff. So the unwanted machine became my desktop, and eventually I set up another box as a server for my (then) website, and as a firewall/router for my friend Bruce. That server has been serving us faithfully for many years, a Pentium I, 100 MHz machine with RAM varying from 32 MB to 80 MB.

Over the last month, however, I started planning for some future projects and my experiments showed that “the little server that could” just “couldn’t” any more. I needed more raw CPU power. So I replaced the server with my Mom’s old Celeron 500 MHz machine, and that’s where my trouble began. Twice, the machine locked up for no reason when reading in from swap during my experiments. It was running the latest FreeBSD 5.4. So I wiped it clean and reinstalled FreeBSD 4.11, the latest stable version from the 4 branch, thinking that somehow the problem was with the 5 branch of FreeBSD. But this morning, doing a simple cp -Rp /oldroot /usr/, it locked up again.

Damn, this is annoying.

I’m starting to suspect the new hard drive, or its interaction with the hardware. And this is where my power as a FreeBSD god fails. I can’t do nothing about how well the system runs if it’s running on flaky hardware. And so I’m frustrated, so frustrated, yet again. I have a few ideas on how to proceed and resolve this whole mess, but for the moment, I’m just waiting on Bruce’s assessment of the situation. Blogging this is part of my therapy to step back and perhaps feel better about the situation. How’s it going so far?

I do admit to one fundamental mistake: running experiments on a production system, or more accurately, replacing a tried-and-true production system with an experimental system. I could have (and should have) run both systems at the same time, ensuring the stability of the new system before retiring the old one. But the one thing that stood in my way was the number of ports in Bruce’s hardware router. I should have just told him to go and buy a larger hub. Oh well, back to work…