Question about RAID set names

Jane Liu janliu at
Thu Sep 23 17:47:10 UTC 2004

Thanks for replying.

> >dmraid -ay, /dev/dm-0 appears again. If I change the RAID set name to
> >"nvraid_blah", then running "dmraid -ay" creates 
> /dev/mapper/nvraid_blah
> >as well as /dev/dm-0. So it seems that /dev/dm-* enumeration is
> >independent of the RAID set names. Does that mean it's OK to 
> change the
> >RAID set name during the lifetime of the set? 
> how did you rename the array?

I changed the __name function in the metadata handler. I am doing this
exercise because the only unique identifier that I can use to name an
array may change with the status of the array. It's a signature that
gets recalculated if the array becomes, say, degraded. My only other
choice is to use the product ID, which is not unique. 

We also have a morphing feature that is currently supported under
Windows, which enables the user to convert one RAID type into another.
Imagine someone has a dualboot system with a RAID0 in it. Then he
converts the RAID0 into a RAID1 under Windows. Let's say all the data
still fits in the RAID1, so the data is basically intact. Now he boots
into Linux, and discovers the array. What would happen to applications
that rely on that data? I would say this is similar to the scenario
where a disk D0 gets replaced with another disk D1. If both D0 and D1
show up as /dev/sda, and all the data from D0 is ghosted onto D1, then
the replacement should be transparent to apps that rely on the data. Is
that right?


More information about the Ataraid-list mailing list