Richard Hughes (hughsi...@gmail.com) said: > In https://fedoraproject.org/wiki/Features/OfflineSystemUpdates we've > implemented doing the package updates at first-boot time. This makes a > lot of the hard-to-fix problems a lot easier. The question then > becomes, how do we make the OS Update process even smarter? A simple > check would be to see if X started after doing an upgrade, and if it > failed, to rollback to the disk snapshot or / (and /boot?) that we > previously knew worked. > > Do do this we can currently use anything-on-lvm, or btrfs and quite a > bit of shell-foo. I'm quite keen on no adding lots of tricky code to > PackageKit to deal with all this complexity, so what about using > snapper? See http://en.opensuse.org/Portal:Snapper for more details. > > It's an OpenSuse project, and other than a small patch I've sent > upstream to get things compiling on Fedora it looks pretty small, > self-contained and sane. > > Does anybody have any better ideas than snapper? I really don't want > to roll my own on this one unless there's a good reason.
If you're using the yum backend, it already has support for this via a plugin. (Haven't tested it recently, but it's been there for a few releases.) snapper's just a wrapper around the other commandline tools, right? We do have 'system-storage-manager' now; see the 'ssm snapshot' command it provides. Bill -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel
