[linux-lvm] Strange LVM2 on RAID Behaviour

[Lucas Barbuto <lucas@stabat.com>]

Hi All,

I'm having a weird problem with LVM2 on top of Linux software RAID-1.
I've installed Debian Sarge recently, using the new Debian Installer and
I've configured it with LVM2 on top of RAID-1.

I'm running the 2.6.8-1 kernel from the Sarge installer with their
initrd.  I've got two 80GB SATA hard drives, partitioned as in [1].
I've got /boot on it's own RAID-1 device (md0 = sda1 + sdb1), becuase
GRUB didn't support kernel on an LV (and still doesn't without patches
AFAIK).  Then I've got md1 (= sda2 + sdb2) and md2 (= sda3 + sdb3)
combined to for a volume group, vg0.  Then I've got md3 (= sda4 + sdb4)
which is used by vg1.  I've created all my logical volumes including
root and swap in vg0.  In vg1 I've got a single LV using up the entire VG.

So here's the problem:  When I boot my system, everything works great,
except that md3 comes up degraded, missing sda4.  dmesg contains the
following:

> devfs_mk_dir: invalid argument.<4>devfs_mk_dev: could not append to parent for /disc

repeated 112 times right at the top (I don't think this has anything to
do with it, but I don't know), and:

> md: md3 stopped.
> md: bind<sdb4>
> md: could not lock unknown-block(8,4).
> md: md_import_device returned -16
> raid1: raid set md3 active with 1 out of 2 mirrors
> md: md0 stopped.
> md: bind<sdb1>
> md: bind<sda1>
> raid1: raid set md0 active with 2 out of 2 mirrors

It doesn't mention md1 and md2.  When I try to hot-add sda4 into md3 I
get more "could not lock" errors:

> lbarbuto@saturn:~$ sudo mdadm /dev/md3 -a /dev/sda4
> md: trying to hot-add unknown-block(8,4) to md3 ...
> md: could not lock unknown-block(8,4).
> md: md_import_device returned -16

If I remove the LV from vg1, I can now hot-add the RAID member device
back in and it happily starts syncing.

So, I'm obviously not an expert on this stuff, but I assume that LVM is
scanning devices on boot and taking over sda4 before the RAID starts.
So I tried adding a filter to /etc/lvm/lvm.conf:

> filter = [ "a|^/dev/md|", "r|.*|" ]

To ignore all but md devices --- this makes no difference.  So now I am
stuck and I don't know what else to try.  Can someone give me some
advice?  This is the third list I've tried posting on, sorry if you've
read this before.

[1] partitions

> lbarbuto@saturn:~$ cat /proc/partitions
> major minor  #blocks  name
> 
>    8     0  117220824 sda
>    8     1      32098 sda1
>    8     2   39062047 sda2
>    8     3   39062047 sda3
>    8     4   39062047 sda4
>    8    16  117220824 sdb
>    8    17      32098 sdb1
>    8    18   39062047 sdb2
>    8    19   39062047 sdb3
>    8    20   39062047 sdb4
>    9     1   39061952 md1
>    9     2   39061952 md2
>    9     0      32000 md0
>    9     3   39061952 md3
>  253     1    4194304 dm-1
>  253     2    4194304 dm-2
>  253     3    1048576 dm-3
>  253     4    4194304 dm-4
>  253     5    2097152 dm-5
>  253     6    4194304 dm-6
>  253     7   16777216 dm-7
>  253     8   16777216 dm-8
>  253     9   24641536 dm-9

Thanks,

-- 
Lucas