Cannot boot server using mdadm RAID1 after creating LVM snapshot

[Roger Hunwicks]

I'm running a test server on Precise to act as a KVM host. The virtual machines are using LVM logical volumes as backing stores. I have created several virtual machines by taking an LVM snapshot of a base image. When I rebooted the server, it no longer boots.

After switching grub to use the console I get messages similar to the ones reported here: http://www.redhat.com/archives/linux-lvm/2012-February/msg00032.html - i.e. "Snapshot cow pairing for exception table handover failed"

The server has 2 2TB SATA drives - partitioned into a smallish sdX1 and a large sdX5. sda1/sdb1 make a mdadm RAID 1 array md0, sda2/sdb2 make md1. Both md devices are clean. md0 contains vgSystem with logical volumes for root, boot and swap. md1 contains vgVMs with logical volumes for all the virtual machines.

I can boot the server of a USB stick and then assemble the md arrays and enable the volume groups and everything seems intact

ProblemType: Bug
DistroRelease: Ubuntu 12.04
Package: grub-pc 1.99-17ubuntu1
ProcVersionSignature: Ubuntu 3.2.0-18.28-generic 3.2.9
Uname: Linux 3.2.0-18-generic x86_64
ApportVersion: 1.94.1-0ubuntu2
Architecture: amd64
Date: Sun Mar 11 11:16:44 2012
InstallationMedia: Ubuntu-Server 12.04 LTS "Precise Pangolin" - Alpha amd64 (20120130)
ProcEnviron:
 TERM=xterm
 PATH=(custom, no user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: grub2
UpgradeStatus: No upgrade log present (probably fresh install)

[Phillip Susi (psusi)]

This appears to be a bug in lvm, not grub.

[Roger Hunwicks (roger-tonic-solutions)]

When it boots it pauses for a long time trying to mount the root volume and then the screen shows

udevd[119]: 'watershed sh -c '/sbin/lvm vgscan; sbin/lvm vgchange -ay'' [264] terminated by signal 9 (Killed)
device-mapper: table: 252:8: snapshot: Snapshot cow pairing for exception table handover failed

The logical volumes all seem to work correctly after you boot into a livecd - i.e. I can assemble the raid arrays, enable the volume groups and then mount the logical volumes - at least the ones that contain actual filesystems - the ones that contain vms I haven't done anything with.

[Roger Hunwicks (roger-tonic-solutions)]

I can install ubuntu-libvirt-host in the USB livecd and then run the virtual machines, so all the LVM volume groups seem to be intact.

I don't seem to have any broken volume groups, but does the fact that the boot process is attempting vgchange -ay mean that the boot will fail if any of the LVM volume groups cannot be started, even if the broken one is not needed to start the system?

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.