I have several lvm PVs on top of mdraid arrays. Those LVs that was discovered in initramfs (because underlying MDs gets assembled), but not activated, still not activated during systemd boot, and systemd drops me to an emergency shell. This is a regression over 2.02.100. I suspect, that something wrong with udev rules. I'll try to investigate this problem futher. Initramfs generated by dracut-031/033.
Reverting "69-dm-lvm-metad.rules" to the same state as it was in 2.02.100 fixes this issue.
(In reply to Alexander Tsoy from comment #1) > Reverting "69-dm-lvm-metad.rules" to the same state as it was in 2.02.100 > fixes this issue. You should be more specific. Like which commit from... https://git.fedorahosted.org/cgit/lvm2.git/log/udev/69-dm-lvm-metad.rules.in
This started from commit 8d1d83504dcf9c86ad42d34d3bd0b201d7bab8f6
This bug at least (only?) affects systems with lvm on MD devices. If PV is a disk partition, then lvmetad activation works fine. Also this bug can be triggered only if initramfs is able to selectively activate LVs, such as sys-kernel/dracut.
Upstream decided to fix this on dracut side: RH bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1011704 Pending patch: https://bugzilla.redhat.com/attachment.cgi?id=805841
CCing dracut maintainers then
@aidecoe: Amadeusz, please apply the patch http://git.kernel.org/cgit/boot/dracut/dracut.git/commit/?id=b874d4b5be4c3b81cf394726c354ef7131b5699d
And this one, as well. :-)
With 2.02.103 only dracut patch would be pending
Patch applied. Thank you, Alexander! +*dracut-033-r3 (06 Oct 2013) + + 06 Oct 2013; Amadeusz Żołnowski <aidecoe@gentoo.org> +dracut-033-r3.ebuild, + +files/033-0005-dracut.sh-harden-host_modalias-reading.patch, + +files/033-0006-ifup-do-not-dhcp-on-network-interface-.patch, + +files/033-0007-lvm-mdraid-Fix-LVM-on-MD-activation.patch, + +files/033-0008-mdraid-module-setup.sh-install-configs.patch, + +files/033-0009-mdraid-module-setup.sh-fixes-for-mdadm.patch: + Applied patches with bug fixes wrt mdraid and lvm. + + It fixes bug #486344, bug #485748, possible bug on PPC and bug related to + net-bonding. + + Thanks to Alexander Tsoy <alexander@tsoy.me> for some of these patches and + generally for help. +