Michael -

-----Original Message-----
From: firebird-support@yahoogroups.com
[mailto:firebird-support@yahoogroups.com] On Behalf Of Michael Ludwig
Sent: Thursday, June 07, 2012 1:12 PM
To: firebird-support@yahoogroups.com
Subject: Re: [firebird-support] Dealing with inserts from multiple
transactions

> One solution would be to brutally serialize the problem out of the

way:

> First do Sales, then do Timeclocks. Get rid of the concurrency, get

rid

> of the race.
>
> Another would be to re-schedule Sale records without matching

Employee

> record for later processing. Timeclock records may create Employees,
> but Sale records may not. Back them up to a separate file or

separate

> table and process them again in due time. Employee records will have
> been created and the problem will have been avoided.

Both of those are good ideas. However, I should have made clear in my
initial post that this was just an example. The real situation is far
more complex - we may have 10 different ETL processes running at
different times with overlapping schedules and no way to serialize
those processes.

I like your idea of moving those records with a problem to a seperate
table and processing them later, but it breaks the rule of success or
failure at the dataset level. I can't process 49k records and
postpone 50 of them and still call that a successful ETL.

Thank you.