Re: OVRDBF strangeness

Hi Craig,

maybe it is worth to check the default values of the OVRDBF command in TST and PROD environment?

Alexei
-------- Original Message --------
Subject : OVRDBF strangeness
From : Craig Richards craig@xxxxxxxxxxxxxxxx
To : Midrange Systems Technical Discussion <midrange-l@xxxxxxxxxxxxxxxxxx>
Cc :
Date : 26 августа 2020 г. 19:48

Hi All,
I've got an odd thing happening with OVRDBF and I'm a bit stumped about
where to look next...

Application A feeds data to Applications B and C.
B and C are effectively two different instances of the same database.

The way it does this is Application A uses SQL Create Alias to create two
tables in QTEMP which point to payload tables in Applications B and C
respectively.
Based on data content, Application A will write to either B or C. This
works fine.

Applications B and C need additional records created on another table and
this is handled by a trigger program on the files written to by Application
A.
The trigger program picks up the appropriate library from the trigger
buffer, sets up an activation group override and runs an SQL Insert. And
then removes the override.

This process has worked fine for a couple of years with thousands of
transactions per day.
In the last few months, the write to Application C from the trigger is
failing as the process is actually trying to write to Application B and
getting a duplicate key error.
It appears as though the override is being ignored.

I'm very confused about this because:
- The process worked fine for a couple of years without issue. Now, it
still works fine for Applications B's transactions but always fails for
Application C's transactions.
- The process still works fine in the TST environments which uses the exact
same programs including the triggers.
- I can see no differences in the files being written to with regard to
authority, level or database relations.
- Application A is a simple batch process which just shunts records from
one application to another.

Has anyone else come across anything like this?
thanks,
Craig