Re: Porting SQR to UNIX

[mreagan@epix.net (Matthew F. Reagan)]

>We have a long running task that runs unattended overnight.  My SQR report runs
>in the middle of this, so it is vital that we control how it starts and stops.
>We have to be able to initiate the report from the client and the overnight
>process must wait until the report is finished before continuing.
>
>Has anybody out there ever done anything like this before?  How did you do it?

Howard,

We use two different techniques to accomplish this type of task.  The first
is a quick and dirty (and platform specific) method, while the second is
platform independent:

1)  Let the SQR program submit the rest of the batch job by using CALL
SYSTEM USING.  This submits the remaining job to the batch queue.  However,
this implementation is not easily moved from platform to platform, since
each operating system has a different JCL/JPL/DCL/Shell script and batch
job invocation.

2)  Let the SQR program write a completion mark (update a field in a
specific job control table in the database).  Another program (on the
machine that will run the remaining batch job) checks the job control table
and launches the remaining work when the completion mark is present.  This
method allows the SQR to run on a completely different host than the other
program(s) in the batch job.  Moving any component in the sequence does not
impact any other job, since they are coordinated by the database table
(which all can access).

Matt...


Oh.  *THAT'S* what the Signature is...Alas, I have no witty comments.  I
musta left them back at the office.