Re: MI program unable to locate job - need assistance

On Tue, May 20, 2008 at 4:53 PM, Simon Coulter <shc@xxxxxxxxxxxxxxxxx> wrote:

On 21/05/2008, at 8:05 AM, Paul Jackson wrote:

Which seems normal

Yes. The pointer is NULL prior to being set then is set to the first
WCBT.

but the based on WCBTBL-SPACE var is all hex zero
which means that the size is set to zero (it should be 16752384) which
causes the tables to be skipped hence the not found option. On the
other systems the size is correctly populated and so the table is
searched normally.

I've not seen that before. I guess it could mean that QWCBT01 is not
used (damaged/skipped/empty).

What does DSPJOBTBL show on the failing system?

What does DMPSYSOBJ QWCBT01 QSYS show on the failing system (only
need the first page)

When was the last time the job tables were compressed on the failing
system? Perhaps arrange to have the tables compressed: CHGIPLA
CPRJOBTBL(*ALL) CHKJOBTBL(*ALL) then IPL. Reset these values after
the IPL. Then see if the program works.

NOTE: Good idea to ensure QTOTJOB, QACTJOB, QADLACTJ, QADLTOTJ, and
QMAXJOB are set appropriately before the IPL.

NOTE: Collect the DSPJOBTBL and DMPSYSOBJ information first then
compress the tables.

While it is interesting to pursue this particular problem I suggest
that altering your approach would be better. Either:
a) Modify the MI program to use the job index to locate the given job
or
b) Use Mark's suggestion of the Interrupt Job API

The advantage of a) is that it will be much faster. The advantage of
b) is that it will work without requiring system-state.

Hey Simon,

By dumping the tables I found out that the table size on the failing
system is stored at position 4117 instead of 21. So by adding a
conditional test on whichever one was not zero I was able to fudge the
program. I also changed the start offset to the entries to x'1300'
(based on the job table dumps).

I found (by dumping the job tables) that another two of our systems
exhibit this behavior.

All Lic and OS versions exactly the same.

I agree that an alternate method would be better, but for now this
works. Thanks for your help!

Paul Jackson