[1639] in SIPB-AFS-requests
Status of disks
???@ATHENA.MIT.EDU (ghudson@MIT.EDU)
Tue Nov 29 17:10:27 1994
From: ghudson@MIT.EDU
Date: Tue, 29 Nov 1994 17:07:15 -0500
To: star-maintainers@MIT.EDU
Okay, this morning, rosebud2 had a 1GB disk (inherited from beacon)
mounted as /vicepb, and a 100MB /vicepa partition.
Now, rosebud2 has a 4GB disk (brought over by dkk this morning)
mounted as /vicepb and /vicepc (each 2GB partitions). Don't put any
data on these partitions yet; we're going to need to bring rosebud2
down later tonight (hopefully) to add the 2GB disk, and we don't want
to create an outage.
The 1GB disk from old-beacon is destined for news, and is currently
connected to sipb-server-1. This is so I can test whether we can
physically move a disk from a Dec running Ultrix to a Sun running
SunOS while preserving the filesystem. This experiment seems to have
failed; at boot time, I got:
sd3 at esp0 target 0 lun 0
sd3: corrupt label - wrong magic number
sd3: Vendor 'FUJITSU', product 'M2266S-512', 2107704 512 byte blocks
I ran "format", chose "SUN1.05" as the drive type (what we're using on
beacon's disk), and entered "label" to label the disk. At that point,
mount fails to work:
sipb-server-1# mount /dev/sd3c /mnt
mount: /dev/sd3c on /mnt: Invalid argument
mount: giving up on:
/mnt
SUN1.05 does not make /dev/rz0c quite the same size as it is on the
1GB drives on the Decstation (it's about 50000 sectors smaller).
There is no other available drive type, and to use an "other" type, I
need to know
- The number of data cylinders
- The number of alternate cylinders
- The number of physical cylinders (1659 from /etc/disktab on Ultrix)
- The number of heads (15? is the number of tracks per sector)
- The number of data sectors/track (85)
- The rpm of the drive (3600?)
The first two parameters must add up to the third. The default value
for the number of alternate cylinders is 2. I tried using 1657 and 2,
and 1659 and 2, and each time, I got:
selecting sd3: <Fujitsu>
No sense error during read
ASC: 0xd0 ASCQ: 0x20
No sense error during read
ASC: 0xd0 ASCQ: 0x20
which probably indicates that those are not good choices.
So, assuming we cannot migrate news by physically moving the spool
disks, we will need two 1GB swap disks in order to migrate news. One
of these is already on sipb-server-2; the other will have to come
from either:
(a) ronald-ann, when it becomes ronald-ann2. This means we
will add a disk back to ronald-ann after the news cutover or the rtfm
cutover, whichever is first. This means a short outage, but no big
deal.
(b) bloom-picayune, after the rtfm cutover. This means we
cannot do the news cutover until after the rtfm cutover.
(c) creative use of expire times, tunefs, and sipb-server-1's
internal disk (which currently has 468MB free).
