A Slight Flaw with the Debian Installer
A few weeks ago, I noticed some really good e-Bay deals on used, Opteron-based servers. I mean, they were going for less money than I had to pay for old Pentium III-based servers just three years ago. So, I decided that maybe it was time to upgrade.
Three of the four machines that I bought were Sun Sunfires, each with a pair of single-core 2.4 Gigahertz Opterons and a pair of hot-swappable SCSI hard drives. Since the drives are rather smallish, I figured I’d just combine each pair into one large logical volume.
On the first machine, I installed Fedora 11. When I got to the hard drive setup page, I chose to go with the default Logical Volume Manager setup. The installer then asked me how many drives that I would like to use for the Fedora installation. I chose both drives, and both were automatically added to the logical volume. No muss, no fuss, and everything worked fine. This is typical of the Red Hat-type installers.
On the next machine, I decided to install Debian 5.0. Again, I decided that I wanted to combine both drives into one large logical volume. When I got to the hard drive setup page, I chose the “Guided” option for setting up the LVM. The next page asked me which drive I’d like to use for the installation. I wanted to choose both drives, but Debian doesn’t provide that option. So, I backed out, and decided to use the “Manual” option for setting up the hard drives myself. That didn’t work either, because the “Guided” option had already created the LVM volume group. (That happened, even though I hadn’t been asked to confirm whether to write changes to the hard drive.) So, I backed out again, and choose to open a command-line session. I then used the LVM tools to delete the volume group and physical volumes. I then decided to go “old-school”, and try setting things up with fdisk and the command-line LVM tools. That way, I could get both drives added to the logical volume and continue the installation.
Only one problem. . .
When I finished, I went back to the hard drive setup page, thinking that my new, manually-configured setup would be recognized, and I’d be good-to-go.
Wrong. . .
Instead, my configuration got wiped out, and I ended up with the original single-drive setup.
Okay, no problem. I figured that I’d just continue the installation, and add the second drive to the logical volume later.
Wrong, again. . .
When the installation completed and I tried to boot from the hard drive, I got an error message about how the logical volume couldn’t be found. Apparently, the LVM metadata got messed up when the installer replaced my manual configuration. So, not wanting to waste any more time on this machine, I grabbed my Fedora 11 CD and installed it, instead. Again, no muss, no fuss. It automatically deleted my mess-ups and created the two-drive logical volume.
On the final machine, I gave Debian another try. This time, I just accepted the default single-drive logical volume, and continued the installation. After the installation completed and I had it booted from the hard drive for the first time, I added the second drive to the logical volume. I didn’t want to mess around trying to learn any graphical LVM utilities, so I just did everything “old-school”. (That is, I used fdisk, the command-line LVM tools, and the command-line filesystem resize tool.) Finally, I did a quick “df -h” to verify that the logical volume was resized correctly.
To be fair, I’m sure that Debian isn’t the only distro that has this kind of installer deficiency. And granted, for an experienced Linux user, it’s no huge deal to add a hard drive to the logical volume after the Debian installation has completed. Still though, you have to wonder. . .
If Red Hat can create an installer with an easy-to-use LVM setup, why can’t everyone else?