Bug#829180: mounts sometimes fail at boot

[Daniel Pocock]

Severity: important
Package: systemd
Version: 215-17+deb8u2

Since upgrading to jessie/systemd, one particular system frequently
stops in the single-user mode password prompt during booting.

Looking at the journalctl output, I usually find that some mount has
failed with a line like this:


/dev/mapper/vg00-name device already mounted or /foo/bar/mountpoint busy


The actual mountpoint is not the same each time the error occurs.

Sometimes there is more than one filesystem that fails to mount.

The system is an NFS server but it doesn't mount any NFS shares from
other servers.

In every case where I saw this happen, the system had been shut down
cleanly before the boot.

When this occurred today, I noticed that the mountpoint was on the root
filesystem (e.g. /foo/bar/mountpoint is a directory in /) so it doesn't
appear to be waiting for any other mount.

The problem always seems to refer to LVM logical volumes.

I previously had a lot more logical volumes on this machine but I've
combined several of them into a BtrFS volume because I thought that
having too many LVs may be troublesome.  Now there are 10 LVs.

It was working fine with over 50 LVs with wheezy, before systemd

Regards,

Daniel