On Sun, Jul 31, 2016 at 10:09 PM, Chris Murphy <lists@xxxxxxxxxxxxxxxxx> wrote: > I'm not sure what kind of snapshot this refers to. I have done two dnf > system upgrades from 23 to 24 on Btrfs, I don't recall seeing any Btrfs > snapshots, but that's a weak statement. The earlier version of this man page refers to Btrfs snapshots. https://freedesktop.org/wiki/Software/systemd/SystemUpdates/ The current man page refers to "update script" as being the thing that should do the snapshot. If the update fails, it's supposed to make sure the /system-update symlink isn't in the snapshot, that way upon reboot the upgrade isn't re-attempted (in a possibly endless loop). And also it should presumably update the bootloader configuration so it boots the unmodified snapshot rather than the failed updated subvolume. Thing is, it's not necessary to do offline updates if you can leverage snapshots. Instead they can be done out of band -> make snapshot, have a chroot or nspawn container do the update work on that snapshot, if it fails delete the snapshot, if it completes update bootloader config. It's failsafe. The user can keep working during the update. And there's only one reboot needed, and only if the update completes, if there was a failure, no reboot is needed at all. Seems pretty useful for servers and workstations. -- Chris Murphy -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://lists.fedoraproject.org/admin/lists/test@xxxxxxxxxxxxxxxxxxxxxxx
