Bug#287415: mdadm: degraded md devices not assembled on boot

Mau mavog at hotmail.com
Mon Jul 24 09:51:09 UTC 2006


Hi Martin

martin f krafft wrote:
> Were the arrays still assembled properly?

Yes

> The reason I think you are seeing these errors is because mdadm tries every drive for every
> array and discards those with wrong UUIDs, so this behaviour is to
> be expected, albeit not very nice, I agree.

Yes, I argued that things were going like that; this is not a real
problem but made me think - probably just like many other unexperienced
people - that something went wrong creating the arrays.

I didn't find any reference to this in the package documentation, maybe
it should be a nice idea to add there an explicit description of this
behaviour... or to find a more silent way to look for UUIDs: partitions
with different UUIDs are simply members of different arrays, right? If
this is true the array can't just be assembled and those messages could
simply be redirected to /dev/null... I mean that maybe those messages
should be visible only if the member partitions were explicitly
specified but their UUIDs are different.

> To debug this, I need more output. Specifically, I need to know
> pretty much exactly what mdadm says during initramfs. To make your
> life easier, try the following:
> 
>   boot with "break=mount" appended to the kernel command line
>   when a shell appears, run the following (the leading dot is
>   needed):
>     . conf/initramfs.conf
>     . scripts/functions
>     scripts/local-top/udev_helper
>     scripts/local-top/md
> 
> I hope I am remembering this correctly. Try it and lt me know.

OK, I'll try this as soon as I can have my hands on that keyboard.

Thank you!

Mau





More information about the pkg-mdadm-devel mailing list