Today, a client needed a Fedora > CentOS migration. I ended up migrating 12 servers with a LOT of configuration.
Mainly, what I did is to re-configuring kickstarts; four of them to be exact. Each of these had specific configurations and a kilometer-long post section.
The main issues were:
- Identifying which packages were not available on the CentOS nor EPEL repos.
- Using ethN nomenclature for the networks (bridges, bonding and all)
- Going from BtrFS to XFS (no subvolumes on the later).
Also, the part command on CentOS’s anaconda doesn’t support partitioning options; so it’s kind of hard to give all the flags XFS needs to work well. This had to be taken to a pre and post sections in order to be done right. The bad part is that it prevented me from clearing all the partitions; which I’d love to do; leaving me to manage the fstab by myself and all the rest of the partitioning/mounting related issues.
On the good side, ALL the rest worked fine. I mean, a lot of things were just transparent; things like: firewalls, SELinux, bonding and virtualization worked “out of the kickstart”; which was good news.
So, it was done in a day or so. I am proud of this and I just wanted to write down the main issues so that you might want to check on these if you do the same.