I’ve usually had some minor, annoying problems when upgrading K/Ubuntu builds. I had much worse issues upgrading Windows versions – including total loss of data – and that still holds today. However, updating to Lucid Lynx 10.04 LTS from Karmic 9.10 on my desktop has been my worst Ubuntu upgrade so far.
Important safety tip – Even though all my data is on a separate /home partition, I still backup all my data to a separate drive before upgrading the operating system. That proved to be very wise in this case, though the underlying issue wasn’t Ubuntu’s fault. When in doubt, always back up! When not in doubt, always back up!
Also, read the Lucid release notes before you upgrade. I know that they are long, but most of the forum questions I’ve seen are answered in those notes. They will help you to adequately prepare for the upgrade to Lucid.
The Good News
First some good news and tips. I hadn’t planned to update on the first release date, but I tried something new. I went into System -> Administration -> Software Sources. On the first tab near the bottom, you can find a pull-down list box to choose your download site:
Simply use that pull-down and click on Other. Then choose a mirror rather than the main site. It will default to site in your home country to make it easy. I chose one that I thought was fairly close and had great bandwidth. That was the ticket!
I started the upgrade using System -> Administration -> Update Manager:
At the top of the window, it informed me that a new version 10.04 was available. I clicked on the Upgrade button, and we were off to the races. The update manager informed me about packages that are installed but no longer supported and generally what it would do. I approved it and the upgrade began.
The download and installation took a little less than an hour (didn’t time it exactly). The download itself only took about 15 minutes or so. At the end, it started to reboot itself as usual. Then…
It produced a messed up 640×480 display with a dialog box that spilled off of the left side of the screen wrapping to come in on the right side. It said that my video card wouldn’t be recognized and offered me three or four choices. I don’t remember them all, but I chose to load a default driver. That seemed safest. I clicked OK, but the dialog appeared four more times in exactly the same place. I know that Canonical put out a new Open Source NVidia driver, so I guessed (incorrectly) that it might be causing some issues. I was getting uneasy…
When it finally rebooted, Grub presented three new Lucid kernel choices! Huh? I chose the newest. Lucid proceeded to its default loading screen, but it again looked like 640×480. Then some text overlaid the screen and claimed that my backup drive was either busy or corrupt, and offered to attempt again to mount it if I pressed a letter (don’t remember which one), ignore the problem if I pressed ‘I’, or fix it manually if I pressed ‘M’. I chose to ignore it, but nothing happened. I quickly became clear that Lucid wasn’t recognizing my keyboard inputs. Dead end. A hard reset and second attempt produced the same outcome.
I rebooted and chose recovery mode. Again I saw messages about a busy or corrupt backup drive, but it loaded my RAID 1 array with no problem. Then it locked up tight. Only the power or reset button on my computer would get me out. I rebooted several times choosing different Lucid kernel options, but no dice. It was a brick. I decided to call it a night and tackle it Friday night.
The First Fix
Given the mess at the end of the update process, I immediately suspected a video issue. I booted the System Rescue CD 1.4.0 and mounted the RAID array to get to the /etc/X11/xorg.conf file. There wasn’t one. There was a backup of an empty one called xorg.conf~. Not a good sign. I deleted the backup file, but the system wouldn’t allow me to copy the xorg.conf.failsafe file to xorg.conf. But I figured getting rid of that bad file might help, so I rebooted again to Lucid recovery mode.
This time, I still received the message about the busy or corrupt drives, but Lucid recognized the keyboard inputs! I chose to manually fix the problem and it dropped me into a recovery shell as root. I navigated straight to /etc/X11 and copied xorg.conf.failsafe to xorg.conf:
cp xorg.conf.failsafe xorg.conf
That worked. Next, I examined /etc/fstab in the nano editor, but didn’t find anything wrong. I also looked at /boot/grub/menu.lst, but didn’t see anything odd other than the multitude of available Lucid kernels. Deciding that my work was done, I pressed Ctrl-D to exit the shell and rebooted normally.
This time, I still receive the reports of busy or failed backup drives, but I chose to ignore. Lucid booted through to the login screen! I logged in and it proceeded all the way to my desktop – a 640×480 desktop. The elements were preserved from Karmic, but the panel elements all overlapped. I tried to get to the System menu to change the resolution, but no dice because it was overlaid with quick launch icons. Trying a back door to load the NVidia drivers, I right clicked on a blank screen area and chose to Change Desktop Background, then went to the Visual Effects tab and selected Extra, but it said that it couldn’t find drivers to enable effects. Not good.
While trying to get to the System menu, I click ever further to the left end of the screen. This unfortunately loaded my previous beta 3.03 of Thunderbird as well as Firefox. Thunderbird immediately downloaded my new mail – about 30 messages from the day. That would later prove very bad due to a lack of attention on my part.
Eventually, I reluctantly removed applets from the task bar until I could reach the System menu. This worked, and I was able to get to System -> Administration -> Hardware Drivers and load the NVidia proprietary drivers. This worked, and I restarted the X-Server with Ctrl-Alt-Backspace, which still worked from my Karmic fix. Yes! I had my desktop back.
Skip ahead an hour or so. I had been using the Thunderbird 3.03 beta. I noticed after a while that it and the Lucid installed version were both present. So, knowing the TBird 3.0.4 would simply pick up where I left off, I executed the new TBird w/o backing up the mail that had downloaded two hours before. Bad move. TBird 3.0.4 somehow found my old TBird 2.0 profile rather than my new TBird 3.03 profile with all my new mail and proceeded to import the mail after I mistakenly approved think it was importing the beta’s mail. Net result was the it wrote a new profile over my old 3.03 profile, wiping out all my mail since about January. Ouch!
Since I intelligently backed up the mail just the night prior, I was able to move profile directories off of the disk except for the one that I copied from backup. I edited TBird’s configuration file (/home/.thunderbird/profiles.ini) to point to the profile that I recreated and all worked fine. So, if anyone emailed me Friday, please resend. To be clear, this wasn’t Lucid’s fault or TBird’s, it was mine for not thinking the situation through before blindly clicking on TBird’s menu entry. I went brain dead for just a mouse click, but providentially it only cost me one day’s mail. That’s why we backup.
Oh, about the hard disk. It was fine. For reasons known only to those who live in the exalted world of kernels, when Ubuntu upgrades it usually reorders the list of drives in the system. The RAID array doesn’t live in that list, but everything else does. In Karmic, my backup drive was sdc. Lucid changed it to sda. Lucid didn’t try to find it at all, just locked up on normal boot and reported it busy/corrupt. After I had the system up and running, I used System -> Administration -> Disk Utility:
Notice in the upper right it lists the device as /dev/sda for that hard disk. Would it have been so hard for Lucid to look for it? Apparently, but past versions did fine. To fix this, I edited /etc/fstab as root and changed sdc to sda for the backup partitions. Reboot showed that worked perfectly. I solved the last Lucid upgrade mystery!
The good news is that after I returned my system to full video and disk operation, almost all of my settings from Karmic were preserved. In fact, you can hardly tell the difference from appearances because I’ve customized it so heavily. This post is getting pretty long, so I will write another post after this one about my Lucid discoveries.
Overall, Lucid appears to have snappier performance than Karmic. They say that it boots faster, but it seems about the same to me. Lucid did a great job preserving almost all but my video settings. It did too good a job of preserving my hard disk settings.
Interim Bottom Line
Canonical really needs to do a better job with these hardware issues. I know enough about Linux to fix them, but your average user would be stumped with a bricked computer. That’s not a good way to evangelize your operating system. Not that Windows is any better, but do we really aspire to that low of a standard? I don’t think so. Mark Shuttleworth expends a lot of time and effort on the interface, and that’s good. But the best looking OS isn’t going to win any friends if it can’t handle 2-year-old video cards and reordered hard disks with grace.
As I said, I will post on the little paper cuts in the upgrade in the next post.