BOOT_DEGRADED=true hangs at initramfs prompt with incomplete RAID0 array

[Brian Candler]

This is an ubuntu 12.04.2 x86_64 server.

It has "BOOT_DEGRADED=true" in /etc/initramfs-tools/conf.d/mdadm

It has two internal drives in a mirrored pair, and 22 SSDs in a RAID0 array.

# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md1 : active raid1 sdx6[1] sdw5[0]
      361195328 blocks super 1.2 [2/2] [UU]

md0 : active raid1 sdx7[1] sdw6[0]
      28303232 blocks super 1.2 [2/2] [UU]

md127 : active raid0 sds[18] sdr[17] sdq[16] sdu[21] sdt[19] sdo[14] sdv[20] sdp[15] sdl[1] sdn[13] sdk[2] sdj[11] sdm[0] sdh[9] sdg[8] sdi[10] sdc[6] sdf[7] sdd[12] sda[3] sde[5] sdb[4]
      11002314752 blocks super 1.2 16384k chunks

unused devices: <none>

The problem
==========

If one or more of the SSDs is not detected, boot hangs at the following point:

--------
md127 : inactive sdr[17](S) ...

unused devices:<none>
Attempting to start the RAID in degraded mode...
mdadm: CREATE user root not found
mdadm: CREATE group disk not found
[ 23.638700] md/raid0:md127: too few disks (18 of 22) - aborting!
[ 23.638818] md: pers->run() failed ...
mdadm: failed to start array /dev/md/SSD: Invalid argument
mdadm: CREATE user root not found
mdadm: CREATE group disk not found
Could not start the RAID in degraded mode.
Dropping to a shell.

BusyBox v1.18.5 (Ubuntu 1:1.18.5-1ubuntu4.1) built-in shell (ash)
Enter 'help' for a list of built-in commands.

(initramfs)
--------

At this point it's possible to type "exit" to continue the boot, but only if you have console access.

How to replicate
=============

This is easy to replicate, by removing one or more of the RAID0 drives and rebooting.

It seems not to matter whether the array is listed in /etc/mdadm/mdadm.conf. Originally it was not, but I tried

    /usr/share/mdadm/mkconf >/etc/mdadm/mdadm.conf

and it made no difference.

Desired behaviour
==============

If I have BOOT_DEGRADED=true then I expect the system to continue booting unattended, not drop to initramfs prompt. Combined with "nobootwait" option in /etc/fstab, the system should complete its booting automatically.

Possibly there is a difference between a working but degraded array, and an array with insufficient devices to start at all. This would include things like a RAID5 with 2 or more disks missing, as well a RAID0 with 1 or more disks missing.

I still want the system to be able to boot under these conditions. This would allow remote testing of the disks, reading the serial numbers of the detected disks, rebuilding of the RAID0 array with fewer disks etc.

Updates
=======

To ensure this was not a recently-fixed issue, I did a full update and repeated the test. The behaviour was unchanged.

The timestamp on the initrd file is newer than the timestamp on conf.d/mdadm

root@ar24-5:/etc/initramfs-tools# ls -l /etc/initramfs-tools/conf.d/mdadm
-rw-r--r-- 1 root root 653 Jun 4 09:22 /etc/initramfs-tools/conf.d/mdadm
root@ar24-5:/etc/initramfs-tools# ls -l /boot/initrd.img-3.2.0-45-generic
-rw-r--r-- 1 root root 14650600 Jun 4 09:47 /boot/initrd.img-3.2.0-45-generic

[Launchpad Janitor (janitor)]

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