RE: V5R2 Quality - Notes from the field needed

> From: Walden H. Leverich III
>
> Something I find interesting, in OS/400 where one job "can't
> possibly effect
> another job" it's two jobs doing the same thing that cause the problem.

Wherever did you get this idea?  With a single-level store, the truth is
that any job can affect any other job.  It's just that OS/400 was written by
systems programmers who know how to handle this sort of concurrency.


> There is a database race condition that exists when multiple
> jobs run the same SQL query at the same time AND that query is
> producing data mapping errors.  Data mapping errors occur when
> either the query is performing an operation whose result may be
> in error, for example a divide by zero, or if some numeric data
> in the underlying file is not valid numeric data.  The race
> condition occurs when 2 or more jobs attempt to get the mapping
> error information at the same time.  Encountering the race
> condition can cause queries to fail and possibly the system to
> stop.
> Additional Symptoms:
> SRCB6000317 SRCB4604491
>
> Problem Conclusion
> This ptf eliminates the race condition by gating access to the
> information needed for the mapping error to one job at a time.

I hope your client has fixed the underlying issue, the data mapping errors.
I can see how this never got caught.  Whoever tested it must have made sure
that queries ran even with data mapping errors, but never thought that once
an error was found to run the SAME query on multiple jobs simultaneously.
The thought would be to fix the data mapping errors before putting the job
in production.

In any event, I find this problem to be very severe and annoying, but
totally understandable and not a sign of any impending quality problems,
especially if the fix was made promptly and correctly.

Joe