Description of problem: Machine always boots into an emergency shell. It seems the problem is that some important components still depend on systemd-udev (not systemd-udevd): # find /usr/lib/systemd -type f | xargs grep -E 'systemd-udev[^-d]' /usr/lib/systemd/system/fedora-wait-storage.service:Wants=systemd-udev.service systemd-udev-settle.service /usr/lib/systemd/system/dracut-pre-udev.service:Before=systemd-udev.service dracut-pre-trigger.service /usr/lib/systemd/system/dracut-pre-trigger.service:After=dracut-pre-udev.service systemd-udev.service /usr/lib/systemd/system/dracut-pre-trigger.service:Wants=dracut-pre-udev.service systemd-udev.service # rpm -qf /usr/lib/systemd/system/fedora-wait-storage.service initscripts-9.39-1.fc18.x86_64 # rpm -qf /usr/lib/systemd/system/dracut-pre-udev.service dracut-020-22.git20120702.fc18.x86_64 Note that dracut is one release older than the latest, but that's because latest dracut is broken - bug 838105. However I'm pretty sure it was complaining about this even before I downgraded. Version-Release number of selected component (if applicable): systemd-186-1.fc18.x86_64 How reproducible: 100% Steps to Reproduce: 1. Boot machine, it fails and drops into an emergency shell. 2. In dmesg: [ 4.088406] systemd[1]: Cannot add dependency job for unit systemd-udev.service, ignoring: Invalid argument [ 4.088452] systemd[1]: Cannot add dependency job for unit systemd-udev.service, ignoring: Invalid argument
OK maybe dracut one is fixed by the latest version, according to the changelog: http://koji.fedoraproject.org/koji/buildinfo?buildID=329329 However initscripts one is still wrong, and the machine isn't booting.
Actually it's not fixed even in latest dracut: # rpm -qf /usr/lib/systemd/system/dracut-pre-udev.service dracut-020-51.git20120706.fc18.x86_64 # grep systemd-udev /usr/lib/systemd/system/dracut-pre-udev.service Before=systemd-udev.service dracut-pre-trigger.service
This seems likely to have been caused by dracut (bug 838367), so close out this bug if it doesn't look like a real one.
The unit naming stuff in initscripts fixed in 9.38-2, anyway.
This appears to have been fixed a while back. Closing.