Updated: January 20, 2012
Just as I thought the storm has passed, there comes another upgrade for CentOS, this time version 6.2. My CentOS 6.1 box is working fabulously after a rather painless procedure that took only about fifteen minutes to complete. The one problem was the Nvidia driver, which was not installed for the latest kernel, so I had to grab it and install it again.
Armored with confidence and experience, I set about repeating the upgrade procedure, only this time, I made sure that Nvidia kernel module was installed, so that it would be included with the rest of the packages. Indeed, this is going to be an interesting test in usability. Anyhow, let's see how well I fared here, moving from CentOS 6.1 to 6.2. Follow me, dear geeks.
Like the last time, I powered up the package manager and let it download the necessary list of updates. CentOS 6.1 to 6.2 entails 310 packages worth almost 430MB of data. No worries, let us proceed.
I got an error. What?
To really understand the nature of this error, I closed the GUI and ran yum update from the command line. yum complained about a duplicate package, namely both i386 and x86_64 versions of the tcp_wrappers being identical. It recommended using --skip-broken flag to get around the problem.
I decided not to do that and explore a little more what this error really meant. First, it was not caused by any conflict with the third-party repositories, so this might something to do with the base repository. After inspecting the contents of these two packages and related dependencies, I realized it was safe to get rid of the 32-bit version. How it got there is another tricky question, but I have no answer for that, especially since the problem did not occur the last time. But we must continue. So I removed the package and reran the yum update command.
This time it worked well:
This time, things were working fine! I had the Nvidia kernel module installed and the correct xorg.conf file in place, so I did not need to repeat the quick installation fix I did in version 6.1. After about fifteen minutes of work, once again, CentOS had been successfully upgraded.
Now, looking at the system clock, you may assume that the upgrade procedure took one hour. You will have to take my word for it, but it was fifteen minutes from yum update to getting the shell prompt back. The reason it seems to have taken longer is that I did take a break and had my first dinner for that day. Yes, I eat two dinners.
Good news, the system usage has gone down even further. Used to be around 320MB and now it's down to just 300MB. By far, the most elegant, fastest, responsive, and resource friendly distribution around.
Yes, you want these self-flattery tutorials:
Like the last time, the upgrade procedure completed with one error. This time, it wasn't the graphics driver, it was one of the would-be broken packages. Overall, I find this second problem to be more serious, as it could easily be any which generic error that common users might not be able to resolve. Moreover, it breaks the upgrade procedure, unlike the Nvidia issue, which lets you continue, upgrade and boot into your desktop with the basic graphics support. Now, not proceeding when faced with errors is a good thing, but the resolution ought to be silent and automatic.
On a scale of 1 to 10, this second upgrade procedure gets 8/10. The main reason for the non-perfect score is the package conflict, which MUST not exist in a bedrock of solid sanity that CentOS represents. However, once past that stage, it's all phenomenal. The distro is as stable and fast as before, probably even faster if that's mechanically possible. Well, then, I guess this concludes my second successful Linux upgrade procedure ever, both times on top of CentOS. This must mean something.