[NBLUG/talk] A new twist on an old RAID problem

[Lincoln Peters]

When I tried to set up a RAID array involving USB drives, it ended up a 
failure primarily because I couldn't get the kernel to wait for USB mass 
storage devices to be detected before trying to set up arrays.  I recently 
upgraded to a custom-build 2.6.19.2 kernel, and I am now experiencing a 
disturbingly similar problem.

It would seem that the codebase for serial ATA controllers has changed so that 
it doesn't stop to scan for drives before starting other kernel drivers.  My 
root filesystem is a RAID-0 array with one piece on a parallel ATA (a.k.a. 
IDE) drive and the other on a serial ATA drive.  However, since those other 
kernel drivers include the "md" (i.e. RAID) driver, that means that the 
kernel tries to assemble and start the RAID array before it has any chance of 
figuring out that there is a serial ATA disk that's part of the array!


I'm pretty sure this is a fairly recent phenomenon, as I didn't have any such 
problems when I built my own 2.6.16.20 kernel.  Has anyone else encountered 
this problem, or found a way to resolve it?  Should I revert to my old 
2.6.16.20 kernel?


-- 
Lincoln Peters		<sampln at sbcglobal.net>

I have had my television aerials removed.  It's the moral equivalent
of a prostate operation.
		-- Malcolm Muggeridge