Re: sqlldr leaves records on the plate...

"J. Drummey" wrote:

> For a migration project we're loading about 20 million records from flat files.
> The records are 436 bytes each.
>
> We've broken up the data into seven files of slightly less than 2gb each and
> run sqlldr(s) concurrently with direct=true and parallel=true.
>
> All records from the first file load without a problem. Each of the next 5
> files loads perfectly except the last 18 records. At the end of operation,
> we are 90 records short ( 18 x 5 ). The last file loads without a problem.
> (It's much smaller ).
>
> Sqlldr gives an error on the first of the 18 records, then
> ignores the rest of the file. If we take the same records from the end
> of each file and start another sqlldr, the records load up fine, so it
> doesn't seem data-related.
>
> I assume we're running into some built in size limitation. For some reason it
> doesn't bother the first file ( which is actually the largest. )
>
> Anyone have a clue?

I'm not aware of any limitation that would do what you are experiencing. My suspicion would be that there is something wrong with that 18th from the last record, perhaps a hidden character, causing it to abort.

Try taking one of your files and editing out the 19th, 18th, and 17th records from the last and see what happens. My money is on the problem disappearing.

Equally. I suspect that if you try to load those three records by themselves it will fail.

Daniel A. Morgan Received on Wed Jun 20 2001 - 12:13:33 CDT