OT: redhat

[Bill Vermillion]

When asked his whereabouts on Sun, Nov 14 05:11 , Fairlight took the 
fifth, drank it, and then slurred: 

> When asked his whereabouts on Sun, Nov 14, 2004 at 12:19:46AM -0800,
> Bill Campbell took the fifth, drank it, and then slurred:

> > >That makes sense.  Does Linux still change the numbers if you
> > >remove a drive in the middle - say you had 4 drives and removed one
> > >- do the drives above the middle one get renumbered.  That used to
> > >be a problem but I'm not that deep into Linux at the moment.

> > I'm not sure what Linux does when a partition is removed, and
> > I really didn't want to find out the hard way which is why I
> > use LVM.

> He said -drives- though. I think he's asking if it will change
> /dev/sdd to /dev/sdc if you remove the original drive that was
> found at /dev/sdc.

At one time it did that.  And of course it's hda and hdb for the 
primary and secondary on the primary controller and hdc and hdd for
the same on the secondary controller.

When you add-in another controller those will typically be the next
in line.  In one of my beasties the only HD is at ad4s - that is
because the primary IDE controllers automatically ad0 through ad3.

> I don't even know the answer to that. I haven't had occasion to
> try. I don't think IDE will downshift, partly because I've seen
> the primary on the secondary bus show up as /dev/hdc when there
> were no HD's at hda and hdb. That's relatively flimsy evidence
> though.

> My guess is that they -will- downshift in SCSI. The only thing
> it has to go by is ID and LUN. The devices aren't mapped
> to those, but rather the lowest ID for a particular device
> (sd*, scd*, st*, etc.). I'm thinking they'll probably shift
> downwards. But then, so does Win9x. If I delete my (unused) D
> partition on the middle drive, all my drive letter assignments
> go to hell in Win95. I keep it there to make things sane. I ran
> into issues with re-mapping the drives in win95, so I had to go
> with this for some obscure reasons.

The MS world has had problems with this since day one.  At least in
XP you can lock the letter to the hardware, and you can re-order
the cds, dvds, etc., to your liking so they don't come up in the
order that MS seem to think is best.  That was one of the poorer
design decisions by MS. {I use he word 'design' tounge-in-cheek'}

And in the SCSI world you can specify which device you will boot
from.  The problem with the early SCSI controllers is that they
mimiced the MS world to boot from drive 0.

Your boot device should realy be ID 6 as priority maps downward
from 7 [the controller] to 0, then from 15 to 8, and 23 to to 16,
and finally 31 to 24.   The saving grace as that none of the
popular OSes had implemented SCSI properly so that's not a problem
in the Intel based world.

I was looking at the Novell manual a friend had for taking her
Novell certification.  It mentioned how SCSI booting from drive 0
except for IBM which did it the wrong way.

In actuallity IBM was the only one who did it correctly.  As much as
anyone decries the IBM ways they do adhere to design specs and
implementations to the letter.  And most of the HD improvements
that have been made - going back for 50 years - have been from IBM.

The only other problems I've read that others have had are the
ordering of the SCSI controllers if you have more than one.

In an old email Bela and I were discussing limits in their EISA
implementation and with the number of controllers the SW supported
and the number of LUNS you could add, theoretically and SCO
system could handle in the neighborhood of 5000 SCSI drives.

This was using 15 controllers, each with 3 busses. So multiply
the busses by the number of IDs by the number of LUNs.

For a wide it would be  15cr * 3bus * 15 IDs * 7 Luns.  Or 
if you do the math 4725 drives.   If you figure a spinud delay of
10 seconds so you don't smoke the power supply if all drives came
on at once the last drive would spin up about 13 hours later.

The basic SCSI design is strong.  If it had been implemented
properly you could have given instructiont to copy a particular
hard drive to a tape drive and then let the OS know when it is
done, all while the OS is doing something else.   

Given some of the SW we've all run across it's probably best that
was never done as users would be saing "I can't access HD number
NN" not realizing they had made it busy doing something else.

Bill
-- 
Bill Vermillion - bv @ wjv . com