Re: raidstart used deprecated START_ARRAY ioctl

[Norberto Bensa <norberto+linux-kernel@bensa.ath.cx>]

Hello Neil,

Neil Brown wrote:
> On Friday July 9, norberto+linux-kernel@bensa.ath.cx wrote:
> > Hello,
> >
> > What does this mean and how do I fix it?
>
> If you have a degraded array, there is at-least and even chance that
> raidstart will not successfully start it for you.

Aha. It didn't start the array when I compiled md and raid0 built-in; that's 
how I discovered this.

It works as modules (weird to me, but I'm not a kernel guru)

> So, you should stop using raidstart.

Ok.

> The options are:
>
>  1/ use "autodetect".  I'm not a big fan of this personally, but it is
>    much more reliable than START_ARRAY.

I already have. But Gentoo uses raidstart (This is pure guessing, I need to 
dig into the init scripts.

>    This is done by set the partition type of all partitions that
>    contain part of an MD array to "Linux Raid Autodetect" (0xFD).
>    Then all arrays are found and assembled at boot time.
>    This requires having all of md (that you need) compiled into the
>    kernel, not as modules.

Did I get that right? Can I get rid of raidstart and the array will be 
"assembled by the kernel"?

>  2/ use mdadm.  Read the man page about ASSEMBLE MODE.
>     You have  an /etc/mdadm.conf that lists
>       - devices (or partitions) to scan
>       - arrays to be started
>       -  UUID of each array
>
>     and mdadm will find and assemble the arrays.

I think I have read something about mdadm in Gentoo docs somewhere. I really 
need to check that out.

Thanks Neil.

Best regards,
Norberto