[comp.sys.nsc.32k] Disk problems

sverre@lars.Seri.GOV (Sverre Froyen) (06/12/90)

I just bought a new disk (Micropolis 1578) to replace a dead one,
however, I have problems bringing it up on my ICM3216.

The drive formats without any errors and I can read and write
to it.  I (re)partition the drive as follows:

Partition table for SCSI device 1:0

partition   first block   last block   number of blocks
    0               2        16385       16384 (  8 Mb)
    1           16386        94209       77824 ( 38 Mb)
    2           94210       276157      181948 ( 88 Mb)
    3          276158       458105      181948 ( 88 Mb)
    4          656438       656501          64 (  0 Mb)
    5          640054       656437       16384 (  8 Mb)
    6               0            1           2 (  0 Mb)
    7          458106       640053      181948 ( 88 Mb)

I can copy the loader and root file system to partitions 4 and 0 and
boot unix (Rev H).  I then proceed to make the other file systems using

	mkfs /dev/dsk/0sx 90974 1 540

for partitions 2, 3, and 7 and 

	mkfs /dev/dsk/0s1 38912 1 540

for partition 1.

After having done this I can observe a number of different
problems:

1) Making the filesystems in the sequence 1, 2, 3, 7,
causes labelit to report that 0s1 has 181948 blocks.

2) Reversing the sequence made the root file system vanish
on one try and made all the blocks in the /usr (0s1) file
system disappear while cpio-ing the usr files in a second.

3) I have also seen the /usr file system being clobbered
after mkfs-ing another file system (possibly 0s7).

I have tried reformatting the drive (no errors again) and rearranging
the partitions with 2 and 3 larger and 7 as the (default) entire drive.

The errors appear to be associated with writing on the last part
of the drive (partition 7).  I have brought it up skipping partition
7 and it appears! to work.

This may, of course, be a bad disk drive or bad cabling, however,
this drive is larger than anything I have tried on the ICM before,
and I suspect there may be driver problems (e.g., if you use more
than ~32000 2^15-1? inodes, df will report "not a file system").
Has anyone observed anything like this?

Since the drive is destined for the pc532 an ICM software problem
is not all that critical, however, if the drive itself is bad
I need to find out now.

BTW, the drive works with the asyncronous SCSI port on the pc532
and gives read and write transfer rates of 700 kB/s and 440 kB/s
respectively (on 3MB transfers).

Sverre

-- 
Sverre Froyen
sverre@seri.gov, sunpeaks!seri!sverre