[Novalug] Kickstart Disk Assignment

[James Ewing Cottrell 3rd]

  On 10/1/2010 11:04 PM, Peter Larsen wrote:
> On Thu, 2010-09-30 at 21:47 -0400, James Ewing Cottrell 3rd wrote:
>> On 9/30/2010 8:29 PM, Peter Larsen wrote:
>>> On Thu, 2010-09-30 at 18:39 -0400, James Ewing Cottrell 3rd wrote:
>>>
>>>> My machine has one internal SATA drive, and I am connecting a USB drive
>>>> (Western Digital passport) to it.
>>>>
>>>> Under CentOS 5.5, it calls the internal SATA sda and the external USB sdb
>>>> Under Fedora 13,  it calls the internal SATA sdb and the external USB sda
>>> The same box or two different boxes?
>> Same box.
>>>> Of course, I want to build the USB drive to actually come up as sda; I
>>>> will select that drive from the Boot Menu.
>>> Since it's two different distros they may very well have different udev
>>> rules to apply.
>> Yes, but they are both "Red Hat". So what changed and why? I did
>> speculate a bit on why...maybe someone knows for sure.
> That may not be as important as just noting it's two different OSes.
> Things change - knowing why isn't changing that fact.
Yes, but knowing why lets me know what to expect.
>>>    There is an option on the "part" parameter called
>>> "onbiosdisk". You may want to use that to bypass what the udev thinks
>>> about the device sequencing?
>> Onbiosdisk is mentioned in the docs, but never specified. What are they
>> and how would I find out what values to use?
> Yup - I couldn't agree more. But it looks like if you know the BIOS
> order, you start --onbiosdisk=80 for the first disk, =81 for the second
> etc.
I figured you might say that, But aren't those IDE disks? What numbers 
are the SATA and USB drives?
>> And no, I didn't change the BIOS settings in between installs! FWIW, I
>> used the grub on the SATA drive to boot the anaconda, and it didn't seem
>> to matter which drive I booted from. Here are two grub stanzas I used
>> for that:
> USB is a much trickier than a standard disk controller. It's not only a
> matter of trying to enumerate the USB ports themselves, but figuring out
> if the USB preceeds the traditional controller in sequence. When the
> boot device changes to USB, the sequencing is up in the air. It may even
> be inconsistent internally. Its odd - definitely. But it's not
> unexplainable.
Yes, I know, which is why you really need to have Only One, because it 
may even be indeterminate which USB device responds first. But SATA 
disks are no picnic either...there are usually SATA ports on the 
motherboard labelled 0 thru 3, but there may be other SATA cards.

For years the algorithm was something like

letter=a
for bus in PATA SCSI SATA USB
     for num in 0, 1, 2, ... n
         if probe($bus, $num)
             assign sd$letter to disk($bus, $num)
             increment letter
end end end

Actually, the PATA drives were called hd, but that's a small detail.

I really don't care what happens....I just want to know what to expect. 
I don't even mind a Rule Change; there may be a Good Reason for it.

I keep copying Paul on this....maybe he will come up with an explanation 
or knows someone who will.

JIM