On Sat, 8 Jun 2013, Steve Langasek wrote:

It's expected that LVM volumes will be activated by udev.  I confess I
hadn't checked whether this works right currently in Debian; while I use LVM
on my Debian systems, my root fs is always on LVM as well, which means the
VG activation is handled from the initramfs before upstart starts up.

Subsequent investigation lead me to the same conclusion (LVM done by udev). /etc/init/udev.conf has "start on virtual-filesystems", but I wonder if it is not emitting that until all the filesystems have completed fsck.

From your fstab, would it be correct to assume that /usr/local and
/mnt/music are the only two filesystems on LVM?  (I ask, because some people
will use UUID in their fstab for LVM filesystems - though this is an error,
as, due to snapshotting, fs UUIDs are not guaranteed to be unique on LVM).

That is correct (of the mandatory ones - not marked user, noauto). I don't quite know why I'm now using UUIDs rather than labels or partition names, but it was stable before.

If so, you should be able to work around this incompatibility by adding a
'nobootwait' option to your /usr/local fstab entry.  This will instruct
mountall to not hold up the boot waiting for this filesystem to become
available, breaking the deadlock between /etc/rcS.d/S??lvm2 and mountall.
This means there will be a slight race condition introduced on your system,
but assuming you aren't starting any services out of /usr/local at boot
time, this should be ignorable.

I will try that. I have no services running out of /usr/local.

If you can confirm this fixes the problem for you, I'll work with the lvm2
maintainers to come up with an appropriate fix here.

I think something (a new job owned by lvm2?) needs to kick off the vgscan and 'vgchange -aly' calls that will result in udev picking up the new volumes.

--
Peter Denison <peterd at marshadder dot org>


--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Reply via email to