I use a file-oriented backup scheme and use 'safekeep' (scripts around rdiff-backup). I'm not in love with safekeep, but it works reliably and is simple to configure. I backup my /home and some shared directories and also /etc from each of 4 machines using safekeep. I do not backup the majority of the rootfs nor /var except for /etc.
My practice is to restore the rootfs and the packages (I keep a package list for each) from the fedora DVD & repo. I keep detailed configuration instructions for the packages on a per system basis, but I also have copies of the /etc directories for reference. In the rare event of a rebuild, I reinstall Fedora from repos, configure the packages in ~90 minutes or a little more. So my approach is NOT a good one for a production facility. OTOH is serves me well.
I have complete file-oriented backups of /home /usr/local/bin and a few others and I can reconstruct these into any partition at any time.
The really great thing is that I have incremental, differential
backups going back ~22 months now. I can reconstruct my /home file system to it's state on any night back to January 2011.
Yes fsarchiver is no doubt a great tool to *copy* partition contents - but it's not really a backup unless you want to keep daily/weekly copies of everything - and that's usually impractical.
Stuff you no doubt know.
There are categorically two ways a backup scheme can save you. The first is when you have a data loss AND you recognize it immediately. Then of course you go to the latest copy and restore. The second case is more common and insidious, you accidentally delete something and you don't recognize the loss until weeks or months later. In this second case a nightly or weekly 'copy' will not help. You need old historical copies going back many months. The only practical way to save that much data is to make incremental backups.
So fsarchive doesn't approach the LT backup issue at all. For example I backup stuff from 4 systems that amounts to ~150GB in compressed form. fsarchive would require the same 150GB every night, ~1TB per week, ~50TB per year. Instead by using rdiff-backup/safekeep I have ~22 months of nightlies available in ~250GB total. So with differential incremental backups I get far more functionality using ~350 times less storage.
So perhaps a one-off nightly restorable image with fsarchiver AND some form of incremental backup together make sense. But let's not confuse the purpose of a restore image with a backup.
Some more germane comments ...