mdadm broken after feisty upgrade, no boot

[Matt Price]

Binary package hint: mdadm

after dist-upgrade to feisty, boot fails with mdadm errors. this is an amd64 system with lvm running over a simple mdadm raid; worked unproblematically in edgy and dapper.

boot errors:
-----

loading, please wait
bogl_init failed: setting screen size: cannot allocate memory
screen init failed
mdadm: no devices listed in conf file were found.
volume group not found
check root- bootarg cat /proc/cmdline
or missing modules, devices: cat /proc/modules ls /dev

alert! /dev/mapper/Anarres--64--main-root does not exist. dropping to a
shell
---------------------
chrooting into the broken system from a dapper livecd, I find that the filesystems are all still intact. trying to run dpkg-reconfigure mdadm givesthis output:

-------------------

W: mdadm: failed to load MD subsystem.
update-initramfs: Generating /boot/initrd.img-2.6.20-6-lowlatency
W: mdadm: unchecked configuration file: /etc/mdadm/mdadm.conf
W: mdadm: please read /usr/share/doc/mdadm/README.upgrading-2.5.3.gz .
W: mkconf: MD subsystem is not loaded, thus I cannot scan for arrays.
W: mdadm: failed to auto-generate temporary mdadm.conf file.
W: mdadm: using the unchecked file and hoping for the best...
update-initramfs: Generating /boot/initrd.img-2.6.17-10-generic
W: udev hook script requires at least kernel version 2.6.19
W: not generating requested initramfs for kernel 2.6.17-10-generic
invoke-rc.d: initscript mdadm-raid, action "start" failed.

---------------------------

my /etc/mdadm/mdadm.conf:
-------------------

DEVICE partitions
ARRAY /dev/md0 level=raid1 num-devices=2 UUID=a1c69140:16931871:f12033d8:e44398ce
MAILADDR root
=--------------------

system is pretty well unusable so of course I appreciate any guidance you can give! thanks,

matt

[Fabio Massimo Di Nitto (fabbione)]

It is a known problem that is being worked out. At the moment the only workaround is to
run a couple of commands from the command line when you are dropped to the shell at boot time.

basically you need to start the raid manually, and then:

lvm vgscan
lvm vgchange -a y
exit

Fabio

PS Yes i am aware this is not optimal but as i said a fix shold land sometimes next week.

[Matt Price (matt-price)]

On Sat, 2007-02-03 at 15:20 +0000, Fabio Massimo Di Nitto wrote:
> It is a known problem that is being worked out. At the moment the only workaround is to
> run a couple of commands from the command line when you are dropped to the shell at boot time.
>
> basically you need to start the raid manually, and then:
>
> lvm vgscan
> lvm vgchange -a y
> exit
>
thanks for the quick response fabio. unfortunately this system has
three lvm vg's, and only one of them is detected by vgscan -- alas, not
the root system, but the vg containing my audio and video storage. so
still no luck!

I don't actually need to use this box for the next couple of days, can
easily wait for the fix, apply it, and see if it fixes my problem. thx,

matt

> Fabio
>
> PS Yes i am aware this is not optimal but as i said a fix shold land
> sometimes next week.
>
--
Matt Price
History Dept
University of Toronto
<email address hidden>

[Matt Price (matt-price)]

On Sat, 2007-02-03 at 16:12 +0000, Matt Price wrote:
> On Sat, 2007-02-03 at 15:20 +0000, Fabio Massimo Di Nitto wrote:
> > It is a known problem that is being worked out. At the moment the only workaround is to
> > run a couple of commands from the command line when you are dropped to the shell at boot time.
> >
> > basically you need to start the raid manually, and then:
> >
> > lvm vgscan
> > lvm vgchange -a y
> > exit
> >
> thanks for the quick response fabio. unfortunately this system has
> three lvm vg's, and only one of them is detected by vgscan -- alas, not
> the root system, but the vg containing my audio and video storage. so
> still no luck!

correction: actually this fix works if the following command is run
first:
./scripts/local-top/mdadm
it may also be necessary to append:
break=mount to the boot line in grub.

>
> > Fabio
> >
> > PS Yes i am aware this is not optimal but as i said a fix shold land
> > sometimes next week.
> >
> --
> Matt Price
> History Dept
> University of Toronto
> <email address hidden>
>
--
Matt Price
History Dept
University of Toronto
<email address hidden>

[harry (harald-dumdey)]

I can confirm that bug on a new installation with the alternate cd (herd2)

running the script ./scripts/local-top/mdadm works...

[Matt Price (matt-price)]

On Tue, 2007-02-06 at 11:06 +0000, harry wrote:
> I can confirm that bug on a new installation with the alternate cd
> (herd2)
>
> running the script ./scripts/local-top/mdadm works...
>

bug is solved by following the instructions
in /usr/share/doc/mdadm/README.upgrading-2.5.3.gz
namely, removing /var/lib/mdadm/CONF-UNCHECKED
and running
update-initramfs -u -k all

which is great for me now, but probably not optimal for a user-centric
distro like ubuntu.

m

--
Matt Price
History Dept
University of Toronto
<email address hidden>

[Bradley Taylor (bradley-railsmachine)]

I confirm this on a dist-upgrade from edgy to herd3. I fixed the unchecked mdadm.conf error following the instructions in the upgrade README mentioned above. On boot, I get dumped into the initramfs and the following works:

./scripts/local-top/mdadm
lvm vgscan
lvm vgchange -a y
exit

My desktop is a 4 disk raid 5 with lvm.

Thanks!

[Rog (roger-mindsocket)]

An alternative for those who don't actually have RAID or lvm2 partitions is to apt-get remove mdadm lvm2 , then run update-initramfs -u -k all

Just a thought. My recently upgraded feisty system still has some major issues, but at least it boots now. :)

[arcterex (alan-ufies)]

The bug is still an issue on herd5, but running the update-initramfs -u -k all on initial boot (after using the 'break=mount' trick) system now boots properly. There was no /var/lib/mdadm/CONF-UNCHECKED though.

Hopefully this can be automated for release :)