I said BOOT, dammit!
December 10, 2008 2:54 pm TechStuffServer’s been down the past few days, since my computer is old and cranky and needs a new set of spectacles to see where its kernel is.
Apparently, older BIOS machines couldn’t access a hard drive with more than 1023 cylinders (most modern hard drives have more; mine has 30401 I think). Usually, this isn’t an issue since the operating system takes care of everything once the drive has booted, as long as the stuff that actually boots the OS is within the first 1023 cylinders of the drive. Initially, everything was fine, but Ubuntu recently downloaded a new kernel (to some spot on the drive that was beyond the first 1023), and when it tried to boot with said kernel, I got an “Error 18: Selected cylinder exceeds maximum supported by BIOS”, and the computer refused to boot. (I had initially attributed this problem to a power outage we had that morning; little did I know that it would have happened the next time I tried to reboot anyway)
The solution to the problem was to create a new partition at the beginning of the drive, throw the boot images on there, and have the OS only reference this “boot partition” when booting. I’m pretty sketchy when it comes to partitioning (my drive only has the main partition, and a small swap partition), and the experiences that I’ve had with partitioning usually involved drives being wiped during the process. Fortunately, I came across some wonderful software called GParted (Gnome Partition Editor) that allowed me to re-partition my drive — get this — without affecting my data. Of course, the partitioning process (which involved resizing my main partition, moving it down the disk, creating a new partition at the front of the disk, then re-sizing the main partition again to fill up the remaining space) took approximately 17 hours of processing time.
After that, all I had to do (after booting with a boot disk; I used the GParted LiveCD) was mount the partition locally, copy the kernel image files over to this new partition, update /etc/fstab to recognize the new partition as /boot (and learning about device UUIDs in the process), and install grub to the MBR (this one had me for a while, since I was installing grub to the new boot partition instead of the MBR of the entire drive, and getting an Error 15 and crashing). Simple! (so simple I just HAD to drag the process out over two days)
Obviously, this post isn’t meant to be a walkthrough solution for this problem, so here are some references in case someone stumbled upon this and has a similar problem:
(and for those in my class that happen upon this site, the stuff I’m supposed to be marking was on my server)

woot :
Date: December 11, 2008 @ 10:40 am
I’m in a similar boat. My server was my desktop in 1998. It’s quite ancient, but has run reliably everyday for its 10 year life so far.
The problem is one day it will fail, and there are not many parts available for it. So, it will be upgraded. The hard part is that until it fails my motivation in moving is simply because it’s easier now than later.
The problem is time mostly. After setting that machine up and tweaking it for all of these years, I don’t remember why I did what I did, nor the extent of the modifications I made at the first boot.
Unfortunately I can’t blame the last guy this time