[comp.sys.dec] HELP!!!! nedded with Fujitsu M2266S drive

sudha@herodotus.cs.uiuc.edu (Sudhakar Yerramareddy) (02/28/91)

We recently installed a Fujitsu M2266S disk ( 1.2 GB ) on our VAXstation
3100.  We used ns:85 nt:15 nc:1658 and made two partitions (a and g).
We filled up about 60% of the 'a' first partition and it seemed o.k.  Then
we were restoring some files onto the second partition from tape.  When this
partition was about 2/3 full, we lost the '.' directory from our first (a)
partition.  We believe that something from the second partition wrote over
that.  We called Fujitsu and Clayton (the vendor) and they suggested that
nc be reduced to 1644.  We remade the file systems, and we had the exact
same problem.  If anyone can tell what the problem seems to be based on the
symptoms and on their experience with this disk, we would really appreciate
it.  If any one has used this disk, please let us know the correct parameters
to use, etc.
Thanks.
-- sudhakar
(for KBESRL, U. of Iillinois at U-C)

grunwald@foobar.colorado.edu (Dirk Grunwald) (02/28/91)

in general, here's how to hook a SCSI disk up to an Ultrix machine.

use rzdisk -g current, record the total number of blocks available.

Invoke mathematica, and use FactorInteger ot determine prime factors
of the number of blocks.

Choose a reasonable number of sectors per track from the prime factors.
Choose a reasonable number of heads from the prime factors.

Divide through and get the number of cylinders.

Bingo, you now have a maximal geometry for your drive. Subtract some
cylinders for UNIX, which still thinks it needs to remap things (the
drive actually handles this), throw the information at a SC/VC spreadsheet
and partition away.

The interface to SCSI drives has no concept of 'cylinders', 'tracks'
or 'heads'. Most drives record a variable number of sectors per tracks
anyway. The UNIX utilities have not kept pace with these drives.

Also, don't forget to enable the cache read ahead on the FUJI drives.
It's a noisy drive, but fast.


Dirk Grunwald -- Univ. of Colorado at Boulder	(grunwald@foobar.colorado.edu)
						(grunwald@cs.colorado.edu)