boot fails for LVM on degraded raid due to missing device tables at premount

[uxeng]

This is a Trusty installation with combined root + /boot within LVM on top of mdraid (type 1.x) RAID1. Raid1 was built with one missing disk (degraded).
[method: basically create raid/VG/LV setup in shell first then point installer at the lvm. At the end of the install create a chroot, add the mdadm pkg, and update-initramfs before reboot.]

The boot process fails with the following messages:
 Incrementally starting RAID arrays...
 mdadm: CREATE user root not found
 mdadm: CREATE group disk not found
 Incrementally starting RAID arrays...
and slowly repeats the above at this point.

workaround:
- add break=premount to grub kernel line entry
- for continued visibility of text boot output also remove quiet, splash and possibly set gxmode 640x480

now @ initramfs prompt:
mdadm --detail /dev/md0 should indicate a state of clean, degraded, array is started so this part is ok

lvm lvs output attributes are as follows:
-wi-d---- (instead of the expected -wi-a----)
lvs manpage this means device tables are missing (device mapper?)

FIX: simply run lvm vgchange -ay and exit initramsfs. This will lead to a booting system.

[uxeng (listmail)]

here is an interesting find:

replacing the missing disk (mdadm --add /dev/md0 /dev/sdb1) and waiting for sync to complete leads to proper booting system
the system continued boot even after I --failed and --removed the second disk.

I could not return the system to it's original boot fail state until I zero'd the super block on the second disk.

some additional messages that I had not seen before (after boot is failing again)

device-mapper: table: 252:0: linear: dm-linear: Device lookup failed
device-mapper: ioctl: error adding target to table
device-mapper: table: 252:1: linear: dm-linear: Device lookup failed
device-mapper: ioctl: error adding target to table
(then repeats the above once more)

[uxeng (listmail)]

attached dmesg output for degraded boot up to premount

[uxeng (listmail)]

attached dmesg output for normal boot up to premount

[uxeng (listmail)]

The primary difference that I see for degraded boot is this error just ahead of activating volume group 'dataone'
'watershed sh -c '/sbin/lvm vgscan; /sbin/lvm vgchange -a y''(err) ' device-mapper: reload ioctl on failed: Invalid argument'

and also after that point the 55-dm and 56-lvm rules do not fire and create the device setup like what happens in the _normal.txt log.

If anyone does compare these logs you can search for the volume group name 'dataone' on both files to see what I am referring to.

[Launchpad Janitor (janitor)]

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

[Mark Thornton (mthornton-2)]

See https://bugs.launchpad.net/ubuntu/+source/mdadm/+bug/1077650