Re: SCSI disk devices

Brian N. Borg (bborg@iaxs.net)
Fri, 09 May 1997 01:06:03 -0500


> Dave Barr <barr@math.psu.edu> writes:
>
...
> > If we do that, we _definately_ have to adopt something like the
> > Solaris /dev/dsk/c0t0d0s3 scheme. It gets really tiring trying to out-
> > guess what the kernel is going to name a new disk when it's put online.
>
> On Mon, 5 May 1997 19:31:44 -0400, Theodore Y. Ts'o wrote:
>
...

> >I had always thought that the right approach would go far beyond
> >scsidev; if you have a volume management daemon that would automatically
> >do the right thing based on the volume ID (pretty much all filesystems
> >have them now: ext2, iso9660, even MS-DOS FAT), then who cares what the
> >device name is?
>

Although both approaches have their advantages, I agree with the latter.
Require a unique "volume label" for each filesystem and use it for the
device name: /dev/dsk/"volume label".

This is the Data General approach in dgux. The only disadvantage is
that
if you do a complete copy, or break a mirror, your volume label is not
unique any more. DG does provide a way out, I believe they also put a
serial number on each slice.

With loadable device driver modules, and pnp, There is no telling what
logical order a controller might end up in.

Under Solaris, you really have to be careful when you start moving
drives
around, less you forget exactly which controller and id number you put
each drive on. And god help you if you move a controller to a different
slot.

--Brian