Bug#684708: mdadm: support external metadata arrays correctly

[Michael Tokarev]

On 13.08.2012 19:40, Dmitrijs Ledkovs wrote:
> On 13/08/12 16:31, Michael Tokarev wrote:
[]
>> Well.  Sort of.  mdmon utility is supposed to be used in the initramfs,
>> with support of taking over, etc.  But this is apparently due to defects
>> in some filesystems (I guess it is ext[34]fs), which performs writes
>> even if asked to do a read-only mount.  I still don't see a reason why
>> a _cleanly_ shut down system should have issues booting - I don't know
>> which case Miquel referred to above when saying that a system does not
>> boot.
>>
> 
> Well... based on the patch, my guess is that because the shutdown does
> not wait for raid to be synchronised the filesystem ends up being dirty
> after a "clean" shut down, hence to reliable boot you should (not must)
> have mdmon with rootfs on external metadata raid.

Nononono.  The _filesystem_ will most likely be clean.  The raid array
is not.

> Can we include mdmon conditionally, if rootfs is on external metadata raid?

There's no reason for that.  mdmon is automatically started by mdadm
if needed.

>> BTW, is it still the case that ext4 performs writes behind the scenes
>> even if asked for read-only mount?  I think I've seen some fixes in
>> this area, but it might be me dreaming.
> 
> True. Maybe we should ask Ted? =)


For this very question - including mdmon into initramfs - i really want
to understand the underlying problem - whenever we're talking about
ext[34]fs which applies journal even for readonly mount and STILL doing
so in 3.2 kernel (wheezy), or something else -- in other words, if
this is really necessary.  I guess it wont _hurt_ to include it (*),
after all, initramfs can sometimes be used as a small rescue system
to modify the real root before it gets switched to.  But I really want
to see what is the problem we're dealing with.

(*) ofcourse it needs to be verified - the takeover after booting
into real system.  I can't do it, since I don't have any spare system
right now to install debian onto imsm array.  I can only verify that
the new takeover does not harm regular boot process without imsm
arrays.

This change - mdmon in initramfs - might be included to wheezy, it
isn't a dramatic change.  But I still don't think it is a good idea
in general -- using native md arrays is a much better way.

Due to the last statement, I'm not sure the severity of this bugreport
is justified.

/mjt