Rational Software Corporation - Trouble Shooting Guide


I don't know the author of this work or even recall exactly how I came by it. I think I saw it on the desk of a co-worker at Transarc and asked for a photocopy, but that's the best I can do.

I have tried to present an accurate rendering of the document as I received it, including a few typographical errors ("problems databases", "prefered", "sofar"). My hard copy has a page break between the "NOW" question and its answer.

Further information about the author or the development of the document would of course be welcome!


[Rational Software Corporation]

                           Trouble Shooting Guide

____________________________________________________________________________

TITLE:          Trouble Shooting Guide
PRODUCT:        Any and All

This guide contains a collection of hints that may help you when
you are facing a problem and you think you tried everything and you
don't know what else to try. The trick is to take one step back and
to look at the problem in a more disconnected way, pretty much like
in "The hitch-hikers guide to the galaxy" where you only saw the
spaceship when NOT staring at it directly but rather letting it
appear in your field of vision from the corner of your eye.

Every problem has its causes and after the problem has been solved,
you always know exactly why you didn't think of the causes at first.
Over the years we have collected these reasons and translated them into
general hints that may help you in any trouble shooting situation.
Translate these hints into causes that might apply to your problem
and try to eliminate these causes.

Check the most frequent causes of problems in your problem domain.
   - is it an access rights problem?
   - is it a problem with case sensitivity?

When/where did this problem NOT occur?
   - what is/was different?
   - can we go back to that situation and forget about the problem?

Is all information on the problem reliable?
   - can you reproduce the problem yourself?
   - is the error message wrong?

Was it asked to do it the "wrong" way?
   - are switches set wrong?
   - read the documentation for the command

Do you have all information that can be acquired about the problem?
   - check all error logs
   - run with the debugger
   - run with logging of (debug) messages completely open
   - build in your own debug messages

Is this a known problem?
   - check release notes
   - check problems databases
   - check personal notes
   - did somebody else run into this problem before?

Does the problem need to be solved?
   - is there a work-around?
   - can some steps be replaced by different steps?
   - can you do something else completely instead?
   - is this problem your responsibility?

Does the problem need to be solved NOW to make progress?
   - work on a different part first and you might stumble across the
     cause for your problem

A problem doesn't need to be understood to be fixed.
   - if you know how the problem can be fixed but you don't know
     why, just fix it

What has helped before solving OTHER problems?
   - reboot machine
   - restart software

Make a list of all factors possibly involved and try changing factors.
   - command, arguments, options
   - product, OS, platform, patches
   - environment variables
   - switches
   - path
   - which user performs the operation?
   - to what is the operation applied?
   - where is the operation applied?
   - access rights
   - machine uptime
   - networking situation (remote user, X-display, NFS)
   - unix processes

Break up the scenario into parts and analyzed them individually or try
similar combinations.
   - if A & B doesn't work, setup separate tests for A and B
   - if A & B doesn't work, try A & C
   - if A & B doesn't work, but A & C works but you don't know why,
     try A & D

Did you analyze the problem both in a top down and bottom up fashion?
   - try to eliminate parts of the scenario while the problem remains
   - think of a small scenario that might exhibit the same problem

Is there a part of the problem you don't like or don't understand very
well and prefered to ignore so far?

Check or change something/anything, even if you don't see why that should
make a change. It might give you additional information to go on.

Explain your problem to somebody else. This forces you to take one step
back and go through the complete problem thinking about every part of it.
In the process, you might stumble across the flaw in your reasoning sofar.

When the problem is solved, store the problem and its solution for future
reference and communicate them to others.

____________________________________________________________________________


Best viewed with any browser
davide+receptionist@cs.cmu.edu