FC2 hd devices swap ids

[Kevin J. Cummings]

Mike McMullen wrote:
> Hi All,
> 
> I had strange occurance involving a hard drive and the cd device swapping identifications.
> 
> I used an old 800mhz e-machine we had laying around to create a backup server
> using rsync. 
> 
> I installed FC2 and updated it to the latest rpms. I created a raid1 (mirroring) device
> for the rysnc backups to live in. Two 60GB drives contain the backup raid device
> (md2). They were hda and hdd. The cd-drive in the system was hdc.
> 
> I got it all configured and everything during testing was working great. I had rebooted
> it a few times and everything came up working great. I powered it down, picked
> it up and moved it to its home in our server room from my office. When it came back
> up, /proc/mdstat showed one of the mirrors kicked out.
> 
> Going through messagges, I noticed that the cd-crive and hdd had swapped designations.
> The cd drive is now hdd and the 2nd disk drive is hdc.
> 
> I used mdadm and added the 2nd mirror back in and it synced up fine. I rebooted the system
> to see if it would swap again and everything came back up fine with the 2nd mirror still
> hdc.
> 
> I thought maybe the cd-drive had gone bad and popped a few cds in to see if they
> worked and they all work as expected.
> 
> Anyone have any idea why the 2nd drive and the cd-drive might have swapped
> designations like that? All I did was power it down and carry it maybe 30 ft.
> 
> It's the first time I've seen this happen in about 10-12 FC2 installs.
> 
> Any input appreciated,

Standard IDE devices get their addresses in one of 2 ways:

	1)  Master/Slave jumper setting on the drives
	2)  Cable select jumper setting on the drives
	    *and* the drives' physical location on the cable.

Sometimes you can swap the devices through the CMOS-Setup, but that is 
usually a motherboard specific configuration.  Are you sure the cable 
connections didn't get 'jostled' during the transport (and are still 
tightly connected), and that the drives are configured "properly" for 
the cabling (master/slave/cable select)?  And that nobody played with 
the CMOS setup settings for the drives?

The only thing I can suggest is that you check the boot logs for when 
the devices get recognized by the Linux kernel and see what they say. 
It may lend some clues as to what happened....

-- 
Kevin J. Cummings
kjchome at rcn.com
cummings at kjchome.homeip.net
cummings at kjc386.framingham.ma.us