Re: Interesting Oracle Data Load Scenario

"Jim Kennedy" <kennedy-downwithspammersfamily_at_attbi.net> wrote in message news:<Nvnrc.34710$gr.3433306_at_attbi_s52>...
> "Matt" <mccmx_at_hotmail.com> wrote in message
> news:cfee5bcf.0405210036.5b3579bc_at_posting.google.com...
> > > Then why not use priority scheduling or resource scheduling in Oracle?
> You
> > > could have the data load log in as a specific user and give them a much
> > > lower priority. (eg not to exceed 20% CPU) Then the load "user" would
> get
> > > no more than 20% of the CPU if all the CPU was being used. If no one
> was on
> > > the system then it would get all of it until somone else needed some.
> > > Jim
> >
> > Sounds like a good idea... I already have a logon trigger defined for
> > ADHOC query users which places them into a lower priority resource
> > consumer group. I could extend this to the data load process too.
> >
> > However as I said in my previous post, the performance is not my main
> > concern. Its the fault tolerance of the load. I want to avoid
> > restoring production in the event of a data load failure.
> >
> > Matt
>
> Even simpler.
> commit;
> do all data load stuff
> if it works then
> commit;
>
> else
> rollback;
>
> If the system goes down in the middle of the load then when it comes back up
> it will rollback the changes. This is exactly what Oracle is good at and
> you don't need to make things complicated.
> Jim

I think the "large amount of data" in the OP might make it complicated. You generally want to start over rather than rollback a large amount.

YMMV. jg

--
@home.com is bogus.
So _that's_ what happened to unify:
http://www.unify.com/products/nxj/jbossthread/jbthread1.htm