The Great Ubuntu Upgrade that Wasn’t
I played with Ubuntu Linux a few days ago and reported that it went on slick and easy and was nice and stable, and I was considering replacing my desktop system with it. Well, in a fit of boredom this weekend I decided to do just that. Because I’m sane, and have had “oops” moments before I decided to do it on a totally separate hard drive so if something did go wrong I could revert back and still get work done. I didn’t have any standard hard drives sitting around, so I decided to install the three 18G super-fast SCSI drives I still had from the old UFies.org system. I later found I also had a couple of 40G IDE drives, but leaving those 15,000 RPM drives and adapter just sitting in my closet seemed like such a crime.
So first of all it was computer surgery time again. I replaced my power supply with the old UFies one as well because it seemed to have way more power connectors. I had a decent power supply in the box already, but it didn’t look like it would support another three power connectors. Then it was man-handling three more drives into the system. Luckily I was able to put them plus the various fans needed without too much strife (or extra noise).
So it began. I first started with the latest desktop release, version 6.10. Seemed reasonable enough. It boots up into the nice sexy live desktop and I click the install button and go through the wizard. Wait a minute, no options for installation onto hard drives that have RAID on them! After a bit of digging I discover I need the “alternative” install CD, which supports LVM and RAID installs. OK, download that and boot up onto that.
Go through and create a drive scheme that looks something like this:
100mb RAID1 for /boot
18 G RAID5 for /
Repeated for each disk. The idea is that /boot would be raid 1 with a spare (not sure what else to do with the extra 100mb of disk on the two other drives) and the rest is basically all RAID5’d, giving me about 36G of system disk, and then my existing /home on an 80G IDE drive would be mounted under the new OS.
OK, complete the install, boot up with the new system. Boot error, can’t find needed files on the RAID partition. Look at the system and find out that somewhere along the way a RAID superblock was put on the main /dev/sda disk, and the new Ubuntu system is helpfully starting that up as /dev/md1, popping my real system disk to /dev/md2. Ok, change the GRUB boot params and reboot. System comes up. It’s nice, works well, which makes me happy. I spent probably about 3 hours fiddling to get the back button on my mouse to work and to get X to display with the proper resolution with my LG L204WT widescreen monitor. The secret I found on a site somewhere (for my own future reference) is to add the following line in the Monitor section:
Modeline “1680×1050” 146.25 1680 1772 1948 2204 1050 1053 1059 1089 +hsync -vsync
OK, finally. Almost ready to “commit” and start using the thing. Then I realized that I was using the “current” release, where as the one I tried a few days ago was the third pre-release version of 7.04, Feisty Fawn (herd-3). Hmmm…. that kinda sucks, why upgrade to something that’s mostly the same as you have now (though I have to admit it did “feel” faster than my current gentoo system somehow).
OK, find a site with how to upgrade, and do it. Basically running a command that downloads and installs everything. Very slick, very sexy.
Next step, reboot.
System comes up with more RAID errors, something about mdadm and the config file not matching or something. Grr….. Ok, Ok, maybe the in place upgrade wasn’t the best way to do it. I downloaded and burnt the Alternative install CD for Feisty Fawn Herd-4 (4 just came out a couple of days ago) and install that. Only problem is, when I’m in the install section of setting up partitions, it is starting up the existing RAID and telling me I don’t have any RAID disks available! ARGH!
After fighting with this for another hour I ended up doing a “dd if=/dev/zero of=/dev/sdx bs=1024 count=10000″ to wipe out any RAID superblock information on the disks. I suppose the problem was mostly because I was using the same partitioning scheme, so I could re-create the partitions and it would lay information down in the same place on the disk, not actually wiping anything out.
After much fighting I finally got it to install. By this time it was about 2am saturday night and I was really tired. So I left it downloading and installing packages and went to bed.
Next morning I was greeted with a happy “click enter to reboot to your new system” message. Yay! Rebooted and….
Same )(&*#$ RAID/mdadm error. ARGH!
<rant>Another thing to note here that because (I presume) of my keyboard setup (ps2 keyboard into an adapter into a USB KVM) the keyboard is active in the POST stage, allowing me to get into the BIOS and once the kernel has booted up and loaded the USB drivers, but not in between. So any time something happened where I needed to get into the GRUB boot loader, or fix an issue booting that happened before I got to a login prompt, I had to crawl under my desk and unplug from the KVM and plug directly into the computer. And then whenever I wanted to swap back to the windows machine to research anything, I’d have to reverse the plugged-in-ed-ness. A PITA once or twice, far more than that the number of times I had to do it :(</rant>
Fought with this some more, trolled through IRC and the forums without any real resolution. Also by this time I was really tired of all the work that was being involved in making life easier for myself, so I reset the BIOS to boot from IDE instead of SCSI and was back in my Gentoo environment.
I actually managed to reproduce the situation in a VMWare session by setting it up the same way, but of course, the second time I rebooted it came up just fine (unlike in real life).
So I think I might wait until Feisty is officially released before I change distros. I considered installing the latest official 6.10 release and then upgrading when 7.04 comes out, but if these issues aren’t fixed by then, I might not want to get to comfortable in the new system 🙂