Bug#653440: linux-image-3.1.0-1-686-pae: Fails to boot since 3.1, apparently in the md subsystem
Frédéric Brière, 2012-01-14 14:21:10 -0500 :
> On Wed, Dec 28, 2011 at 11:33:03AM +0100, Roland Mas wrote:
>> called triple-patte. The rest of the hard disks is used by a third
>> partition, and /dev/sd[ab]3 make up a two-device RAID1 array called
> Are both members of this array marked as writemostly, by any chance?
> (You can check with mdadm -D.)
Good point, I should have mentioned that. No, this particular array
has no writemostly devices; the other two arrays each have two
writemostly components (the hard-disk partitions), and one “read-write”
component (the SSD partition). However, although this is what I see in
/proc/mdstat, I seem to recall that it is not exactly stored on-disk,
since when I do a full reboot (rather than a hibernate/resume cycle) the
three-way arrays all have three writemostly components and I need to
“echo -writemostly > /sys/block/md127/md/dev-sdc1/state” (and ditto for
the other). So you're probably right.
The irony is, I was looking forward to a 3.1 kernel precisely because
it is said to make the writemostly state persist on disk :-)
> I think you were hit by d2eb35a, same as me; I'll either comment here or
> file a new bug, depending on the case.
I think you're right.
Late frost burns the bloom / Would a fool not let the belt / Restrain the body?
-- in Good Omens (Terry Pratchett and Neil Gaiman)