problems with Interactive UNIX

[Gary S. Trujillo]

The following is being posted on behalf of a neighbor without net access.
Please post replies or send them to me (I will summarize if there's interest).

I hereby make the standard disclaimers regarding being sorry to bother folks
if these subjects have been discussed recently.


							6/23/89

HELP!!

This note describes an NMI (nonmaskable interrupt) trap that causes our
Everex 386 20 MHz Model 3000D computer (with expanded memory totalling 5MB)
to crash during or immediately after statistical calculations in STATA,
which runs on the UNIX partition of our 150MB hard disk.

Hardware:  standard AMI 386 BIOS, standard 32 bit RAM card with 32 bit RAM,
standard controller, standard ESDI hard drive; all conform to IBM standard.
Our version of UNIX was written by Interactive Systems Corp.

The problem:  When STATA runs on a 3MB data file, it often crashes during or
right after a statistical procedure with the following messages:

WARNING:  User NMI trap:  unable to determine source of NMI

PANIC

Attempt to enter Debugger with no db module configured!

continuing...  trying to dump 1248 pages
...........................................

warning:  Kernel NMI trap unable to determine source of NMI

This lasts for several minutes, accompanied by a very shrill bell.

Incidentally, this problem occurs ONLY on files of about 1MB or more, and it
occurs only on about 1/3 of the statistical procedures we try on 3MB files.

     An additional problem, which may or may not be related, is numerous
soft errors on the DOS partition of our hard disk.  A chkdsk revealed that
almost 2MB of our 30MB DOS partition was lost clusters, in 943 clusters.


The following ideas have been ruled out:

1. Bad memory chips; all memory chips have been swapped.

2. Video card:  we have a monochrome monitor, so that can't be the problem.

3. Bad system board:  motherboard has been replaced.

4. Bad PAL chip:  PAL chip has been replaced.

5. Bad 80387 math coprocessor: problem recurs (a little sooner) even when
   coprocessor is removed.

6. Bad keyboard:  keyboard has been replaced.


Outstanding Questions:

1. Are the problems related?

2. How can we run STATA on large files without crashing?

3. A software program called Disk Technician cleans DOS hard disks of
   soft errors.  Is there a program like that that runs on UNIX?
-- 
	Gary Trujillo
	(harvard!wjh12!gst)