Record count discrepancy.

ron

ron

Member
Hi ...

I've come across a problem that has me stumped. I hope someone can help me!

I have developed a system for doing "unattended" dumps/reloads at remote sites. There are a little over 200 sites - each using Progress 9.1D ("SP00") on Unixware. Yes - I know it's all "really old". I've made my comments - but it isn't "my" system.

At most sites the system works fine - but at some sites it aborts due to a record count discrepancy. At the very start of the process it produces a Progress tabanalys report - and it does the same again at the very end. The two reports are cross-checked (programatically) to make sure each table ends-up with exactly the right number of records.

When I examine the logs I find that for one table (occasionally two) the number of records in the first report differs from the number of records dumped. I have eliminated the possibility of any programs "unintentionally" updating the DB during the dump. And - the problem can be reproduced.

I introduced a preliminary step to re-index the table (or two) with the discrepancy - expecting that that would remove the discrepancy. But it doesn't change a thing.

Can anyone shed any light on this problem?

Thanks,
Ron. :(
 
It's a 9.1D thing.

I don't recall the bug#, or even if it is in the kb, but it is a "known bug".

I also don't recall which report is right.

I vaguely recall that a record count based on FOR EACH will come up with the same values for both the new and old databases.

There is also some potential during a dump & load or an index rebuild for certain bogus "duplicate unique index" records to be removed. (Certain bugs sometimes allow these keys to be created...)
 
Thank you VERY much, Tom. That is certainly helpful.

I now have an explanation - and another reason to get Progress updated to a later release!!

Cheers,
Ron.
 
You must log in or register to reply here.
Back
Top