Bug#644876: initramfs-tools: Boot failure from software RAID1 + LVM2 by timing

[Jort Koopmans <jort.koopmans@gmail.com>]

Hi Martin,


First of all, thanks for helping me out.

On Thu, 2011-10-13 at 11:10 +0200, martin f krafft wrote:
also sprach Jort Koopmans <jort.koopmans@gmail.com> [2011.10.12.2143
+0200]:
> > Common guys...can somebody look at my bugreport?
> 
> We have. Let me offer some suggestions:
> 
>   - two days in FLOSS time is not a long time, please be more
>     patient. Or get involved!

Sorry for that, I'm trying to get involved though.

> 
>   - please try to properly format your e-mails to make lines no
>     longer than 68 characters.
> 

I've noticed the longer than 68 chars in my first msg, but it was
formatted by reportbug (?).

>   - if you are criticising initramfs/mdadm, then it helps to
>     reproduce the output you are seeing, ideally after set -x.
> 

True, since this machine does not have a serial port I haven't been able
to log the output yet (but i'll look into that).

>   - we are not common guys. I think you meant "come on" ;)

Roger that ;) 

On Thu, 2011-10-13 at 11:06 +0200, martin f krafft wrote:
> also sprach Jort Koopmans <jort.koopmans@gmail.com> [2011.10.12.2143 +0200]:
> > In the mean while I've checked another solution; moving the call
> > to the local-top scripts till after the ROOTDELAY loop (within the
> > local file).
> 
> init-top/udev also uses it.
> 

Ok, so calling them twice in the local script wouldn't hurt then.

> > This works in my config. But this would delay finding the ROOT dir
> > in normal instances (where devices are quick enough)
> 
> So?
> 
> Instead of patching this here and there with band aids, I suggest
> that everyone with an interest instead invests time in testing
> mdadm/experimental, which provides event-based assembly, and helps
> porting the changes to current mdadm (since I don't have the time at
> the moment). And then, LVM is up next.
> 

I agree that it is a little patchy this way, although the *right* way
you mention will take considerable amounts of time to complete (I
guess?).
I will have a look at mdadm experimental though and see if that works
(just installing experimental on this box and trying the original
initramfs scripts. Indeed LVM is needed too but that's the next stage
(I'd rather use experimental mdadm w/o LVM than a custom initramfs).

Best regards,
Jort Koopmans