How to install a Wren-V on a Sun 4-260

[mccrory%almond.hepnet at csa3.lbl.gov]

I am a physicist at the Fermi National Accelerator Laboratory in Illinois
and I have just finished a research project which I feel is worthy of
reporting in some technical journal.  SunSpots seems to be rather generous
about article acceptance, so here is the results of my research. 

Installing a Wren-V external disk on a Sun 4-260 proved to be quite an
educational experience.  I learned a lot about Sun, SCSI and Unix in the
process.  I present here a menu of how to do it.  I started on Dec 21,
took two four-day weekends to think things over, have Christmas & New
Years, and I finished on Jan 2.  I worked for five days, or three weeks,
or two decades on this project!

1. Buy a Wren-V disk in the externally powered box.  Note that there is no
place to put a true SCSI disk inside a 4-260 box, so you gotta put it
outside.  I spent $2100 for the disk and $200 for the box.

2. Backup the internal disk using dump.  Dont use tar or gnu tar because
you'll probably, at some point, be working from the mini-root. 

3. Strip off all the cute paneling from the 4-260 box.  RFI CAUTION: Your
am radio might not work anymore.  (Big deal!)

4. Make a new 50-wire ribbon SCSI cable.  This cable will allow you to tap
off the SCSI between the disk interface (Emulex MD21) and the tape
interface (Emulex MT02).  The lengths of my cable are approx as follows:
VME to MD21, .75 m; MD21 to Wren-V 1.5 m; Wren-V to MT02, 1.5 m.  I
shrouded the cable in kapton tape to protect it from scratches.  I routed
the cable down through the VME cage and out one of the slots in the back
(who cares about a little rfi?).

5. Remove the terminators from the Wren-V (documentation booklet is
correct on this one!).  Leave the terminators on the MT02.

6. DO NOT connect the new SCSI cable to the MD21 card (for now).

7. Our Wren-V recognized itself only as SCSI address 0, independent of the
dip-switch settings.  So be it.  (This can be determined in the format
step, below, by toggling the SCSI address dip switch on the back of the
Wren-V box, power-on reset, and re-running format.)  Next, change the dip
switch on the lower left corner of the MD21 from 0x00 to 0x01 (LSB on the
left); the old internal disk is now called sd2.  (Disk naming `convention'
can be found, among other places, in the 4.0.3 upgrade manual, Revision A
(7 July 1989), Chapter 4, page 59.)

8. Boot from tape (MUNIX), run format (under SunOS 4.0.3, 4.0.1 and prior
might not be able to handle this large disk), select the Wren-V entry and
partition as desired.  I created `root,' `swap,' and `/usr' partitions to
mimics the original disk.

9. Run newfs on each partition.  If you have booted from the mini-root in
the swap partition of the Wren-V, then you will be `hosed' when you run
newfs on the swap partition.  newfs reports that I got a formatted disk of
568 Mb.  We think we might be able to squeeze a bit more out by
re-partitioning more carefully.

10. Restore `root' and `/usr' onto the Wren-V.  Run fsck on these
partitions. Run /usr/mdec/installboot for the Wren-V.

11. Power down, re-connect the internal disk (now at SCSI address 1, SunOS
disk sd2).  Reboot from Wren-V.

12. Run `/dev/MAKDEV sd2' to create entries for the old disk, now called
sd2.  You can mount the old disk and copy over files you need.

13. Run installboot for the internal disk.  This should allow you to boot
from from the internal disk (>b sd(0,8,0)), but I haven't been able to get
this to work, yet.

14. Have a beer. (Can't stop on step 13!)

Observations: The local Sun support said it probably couldn't be done and
that it certainly wasn't supported, although they tried to help out.  I
did not try at all to get help from Wren (CDC?).  The sticking points
were: (1) the Wren-V only recognized itself as SCSI address 0; (2) it
seems to be important to install the boot block only after you have
successfully installed the root and /usr partitions to the disk; (3) I
kept hosing the disk by making dumb mistakes (e.g., rm -rf /usr; whoops!)
and doing things out of order.

I welcome inquiries on this matter.  Hey! I've nothing better to do!

Dr. Elliott S. McCrory
Fermilab/MS 307
Batavia IL 60510
mccrory at fngate.fnal.gov
(or) mccrory at fnal.bitnet
(or) mccrory at adcalc.fnal.gov
(708) 840-4808