Re: sqlldr and discarded records

"Sybrand Bakker" <gooiditweg_at_nospam.demon.nl> wrote in message news:p2527voicdfo795iluc1ibggeft83qo1kv_at_4ax.com...
> On Thu, 13 Mar 2003 21:30:16 GMT, "FC" <flavio_at_tin.it> wrote:
>
> >If you are loading data into staging tables, the load failure could
trigger
> >an action deleting the partially loaded data or taking some other
"clever"
> >action.
>
> The normal PMON cleanup applies. No 'load failure trigger' whatever it
> may be, required, and such functionality is likely going to stink,
> because it never will address all possible situations.
>
>

I suppose you are talking about uncommitted data. Any other records that were loaded and committed so far are going to remain in the table, no matter why SQL*Loader stopped.

If you have a file with 1000 records, ERRORS=1 and ROWS=100, if an error occurs at record 102, then you will find the first 100 records in the table and possibly an extra one because SQL Loader is committing the extra record before exiting with a return code of 2 (in Windows), as I've seen this morning, while simulating this situation. If SQL Loader is killed abruptly instead, the number of records in the table depends on the number of succesful commits done so far which should be a multiple of the ROWS parameter.

The kind of "trigger" I was talking about is meant to address situations like "maximum number of errors reached", certainly not meteorites or kamikazes blowing up the server.

Regards,
Flavio Received on Fri Mar 14 2003 - 07:40:06 CST