[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Yaird, grub and d-i should also support dmraid devices.



Alle 14:46, domenica 6 novembre 2005, hai scritto:

> > Actually, I'm not sure where this should go. You can also make a case
> > that it should go into hw-detect's disk-detect.
>
> Ah, you may be wrong, i have only little understanding of what dmraid
> really does, and if it is only used to detect the existing partitions
> created at the bios-level, or if it can be used also to create and modify
> them. 

I think Frans is right. BIOS doesn't let you partition your array  but only 
choose the type of it between mirror, stripe and linear addition, you have to 
partition it manually with usual tools.

> I assumed the second and thus suggested partman, maybe both cases 
> should be handled.

I think a question like "nvidia raid array found, would you like to activate 
it?" will be enough. Then we must hide the component devices of the array and 
go on with unmodified partman.

To detect an array we just look at the output of "dmraid -r -c", it gives 
something like "nvidia_jbdaccab" if it finds a dmraid array or "No RAID 
disks" in the opposite case.

> > Also, please look very carefully at which architectures need to support
> > this. I see dmraid is built for S/390, but I very much doubt it makes
> > sense to support it in d-i for that architecture...

This is true, for the moment dmraid makes sense only on i386 and amd64 archs, 
but this can change because dmraid could support any software raid on 
motherboards, having the right metadata.

> I believe it is x86 only, or at least for those arches able to run the x86
> bios in those cards. I may be wrong though.

It is not tied at dmraid internals, but commercial ones I think.

> > Last question: how mature and safe is dmraid? What risk do users run when
> > using dmraid? The package description [1] raises some doubts...

This is a good question.

> Yeah, but is this because it is unsafe to use, because the description has
> not been updated in a long time, or just because the dmraid maintainer
> wanted to take no risk.

This is a good answer.

-- 
ESC:wq



Reply to: