[CentOS] Re: Problem selecting installation drive : 5.2

Sat Oct 25 14:47:33 UTC 2008
Alex H. Vandenham <alex at avantel.ca>

After much frustration and wasted time:

Solution:  Additional Boot option (undocumented) : nodmraid

On Friday 24 October 2008 06:41:07 pm Scott Silva wrote:
> on 10-24-2008 7:25 AM Alex H. Vandenham spake the following:
> > When doing the 5.2 installation the ONLY drive option provided is
> >
> > mapper/nvidia_cbjcdhfe (250G)
> >
> > My system has 2 SATA - 250G drives and I  want to use them in a RAID/LVM
> > configuration.
> >
> > My question:
> >
> > How do I get the installer to let me use DiskDruid to create RAID1 arrays
> > and then use LVM for the mount points?
> >
> > What I've done so far:
> >
> > I've verified that both drives are recognized by going to the shell
> > screen and using fdisk to access both /dev/sda and /dev/sdb - so I know
> > they are both available.
> >
> > I've also tried using the Centos 4.7 installer where I can do exactly
> > what I want.
> >
> > I've accepted the default and after a full install of 5.2 there is no
> > raid and total space is 250G.

the default install DOES use raid but not with Software raid so /proc/mdstat 
shows no devices.  How nvidia drivers/Linux kernel modules handle the raid 
(and raid failures for example) remains unknown - info Anyone?

> >
> > I've looked at the various deployment guides and installation guides but
> > can't find how to do this . . . .

Chapter 8 does not mention the nodmraid option.  When that option is used, the 
two SATA disks are listed as expected.

> >
> > puzzled
> >
> > Alex
> > ===
>
> The default install doesn't do a software raid, and if you are trying to
> use an Nvidia onboard raid controller as a raid device in linux, you are
> probably out of luck.

Apparently the nvidia drivers and/or the kernel LVM modules are creating a 
raid array - hard to know for sure without some documentation to help me 
confirm that . . . 

>
> To do software raid  with LVM over it, you have to do it all manually.
> First creating the raid devices, and then adding the resulting md devices
> as LVM partitions.

Even when I created working Software RAID1 arrays with mdadm (with Centos 4.7 
rescue), on reboot the SATA drives were used by the nvidia drivers (not 
sure ?) to create the array with LVM VG/LV, and then md could not start the 
software raid1 arrays.  

Cheers;

Alex
====

-- 
This message has been scanned for viruses and
dangerous content by Avantel Systems, and is
believed to be clean.