Planning Saves the Day when Upgrade Fails

Not everything goes as planned, even Linux upgrades. I had a narrow window in which to perform an upgrade to a server this morning and some problems arose which forced me to abort the upgrade.

For many upgrades pretesting with virtual machines can tell you a great deal about whether an upgrade will work and how many problems you can expect to have. Working with real hardware is even better.

In the case of this upgrade I did both. But it was really more than an upgrade. Due to the number of versions between the old version of Fedora and the new one, I would basically have to wipe out the old installation and start over. So I decided to simply build the new system on a spare hard drive on one computer and transfer the completed and tested installation on the new hard drive into the old computer. I even preconfigured many of the configuration files that would need to change when I installed the hard drive in the old computer so that all I would have to do is copy the revised configuration file over the original or make a few minor modifications to the files already in place.

I figured I could make a last minute backup, replace the hard drive and make the necessary configuration changes, then restore the files from the backup and resolve any problems within the two hour time window.

This is always good in theory but it seldom works out as planned.

Of course there were problems and I always count on having unexpected problems. In this case the hardware incompatibilities with the new configurations were greater than I could resolve and still accomplish everything else that needed to be done within the window.

My plan for backing out was simple. Put the original hard drive back in the computer. I did this and everything came back up and worked fine. I now have plenty of time to determine how best to resolve or circumvent these issues the next time.

Proper planning is essential to performing any upgrade or change in a production computing environment. This includes the need to plan for a failed upgrade as happened this morning.