I came across a 'small' issue today when writing the policy for our CentOS servers. It was about how to safely run yum upgrade.. and what happens if something goes wrong.
I have daily backups of the server, so should the worse come to the worst, it's easy enough to recover... but I was looking for a smarter way.
RPM Rollback seemed to be the answer, however, for the life of my I couldn't get it to work on my CentOS system.. and as it should have been running since the start, it wasn't really an option.
What I needed was some kind of system that wrapped up the current rpm binaries, with the configs, and timestamped them so that I'd be ready to revert back to a certain timestamp. RPM rollback would handle this theoretically.
Luckily, I have multiple levels of redundancy built into the systems, so even if 'yum update' does bring down my machine, it'll have little impact other than irritation, and frustration at a waste of time.
Therefore, for now I've simply backed up all the rpms from /var/cache/rpm/*/packages/ into /opt/packages.070808/ ; then ran 'yum update.'
There's got to be a better way; please let me know if you know it!