[systemd-devel] 'udevadm settle' brakes lvm on top of imsm raid

[Oleg Samarin]

Hi!

I have an imsm raid-1 device /dev/md126 assembled of /dev/sda and /dev/sdb.
I have a lvm group on top of /dev/md126p2 with some logical volumes. All
this work fine with Fedora 21.

I'm trying to fresh install Fedora 22 in some of lvm logical volume. I boot
with Fedora USB live media and run "Install to hard disk". But anaconda
does not see any existing lvm volumes so I can not choose them as a
destination.

I've created a bug https://bugzilla.redhat.com/show_bug.cgi?id=1178181 on
this issue.

After some discovering I found that the reason of this issue is that
anaconda brakes lvm:

before launching anaconda pvdisplay reports there are two physical volumes
/dev/md126p2 and /dev/sdc2 (/dev/sdc is an SSD that does not belong any
raid and contains a separate lvm volume group)

after launching anaconda pvdisplay reports another two physical volumes
/dev/sdb2 and /dev/sdc2, that is wrong, because /dev/sdb is a part of
/dev/md126 raid.

journalctl shows:

May 28 02:44:08 localhost systemd[1]: Stopping LVM2 PV scan on device
259:1...
May 28 02:44:08 localhost systemd[1]: Stopped LVM2 PV scan on device 259:1.
May 28 02:44:08 localhost audit[1]: <audit-1131> pid=1 uid=0
auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0
msg='unit=lvm2-pvscan at 259:1 comm="systemd" exe="/usr/lib/systemd/systemd"
hostname=? addr=? terminal=? res=success'
May 28 02:44:09 localhost kernel:  sde:
May 28 02:44:09 localhost kernel:  sda: sda1 sda2
May 28 02:44:09 localhost systemd[1]: Starting LVM2 PV scan on device 8:2...
May 28 02:44:09 localhost kernel:  sdb: sdb1 sdb2

Seems lvm stops using /dev/md126p2 and starts using /dev/sdb2 at this moment

I can see in anaconda program.log that it ran 'udevadm settle' at this
moment:

02:44:07,141 INFO program: ...done [9] (exit code: 0)
02:44:09,290 INFO program: Running... udevadm settle --timeout=300
02:44:09,305 DEBUG program: Return code: 0
02:44:09,306 INFO program: Running... udevadm settle --timeout=300
02:44:09,315 DEBUG program: Return code: 0

So 'udevadm settle' breaks lvm from using '/dev/md126p2'. Futher lvm
rescans force it to use /dev/sdb2 instead.

The attached storage.log contains a trace of udev events of this. All other
logs are available
in https://bugzilla.redhat.com/show_bug.cgi?id=1178181

What is wrong? How to force lvm to use /dev/md126p2 instead of /dev/sdb2
after starting anaconda?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/systemd-devel/attachments/20150528/86d736ce/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: storage.log
Type: application/octet-stream
Size: 237365 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/systemd-devel/attachments/20150528/86d736ce/attachment-0001.obj>