Re: Redo killer

mmoore_gmp_at_yahoo.com (Mike Moore) wrote in news:43c454d1.0410251150.21c990eb_at_posting.google.com:

> Part of our billing process creates over 1 million inserts into a
> table which kills our redo logs. Then when the process is complete
> the data is extracted and never referenced again. We truncate the
> table each night. In other words the data is semi-temporary.
>
> I'd change the table definition to a temporary table, except that if
> the program crashes it needs to pick up where it left off and not
> start all over again. There has got to be another way to handle this
> process, but I'm at a loss for a solution. Any ideas would be greatly
> appreciated!!!

Use direct-path inserts.

alter table t1 nologging;

insert /*+append*/ into t1 ....;

For the app to be able to restart you will need to commit regularly and make sure the app is smart enough to pick up from the last commit point.

See chapter 19 (http://tinyurl.com/6cnkd) of the Oracle 9i Database Concepts manual for more details. Received on Tue Oct 26 2004 - 09:17:41 CDT