Look into alter table tablename nologging before your update then alter
table logging before and after your update.
"Eduardo Tavares" <ejtavares_at_tmn.pt> wrote in message
news:Xb699.428$v34.628804_at_newsserver.ip.pt...
> What i usually do when i havo to manipulate tables with such amount of
> records is to slice the updates in smaller pieces ...
> for example you can update 10000 rows and then commit (you can control
this
> with rowid ...) that way you dont need to worry about rollbacks ...
>
> The add column is instantaneous, the problem is the default clause ( wich
> makes an update ... and commits only at the end ... you'll need a lot of
> rollbacks )
>
> Hope this helps ...
>
> Eduardo Tavares
>
> "Peter Laursen" <ptl_at_edbgruppen.dk> wrote in message
> news:3d64c91c$0$43947$edfadb0f_at_dspool01.news.tele.dk...
> > Hi,
> >
> > System(s) 8.0.5 and 8.1.7 SE
> >
> > Adding a new column to a large table is taking hours. Its part of an
> update
> > that will be run at many customers. The table will contain approx 100M
> rows.
> > The table has 5 number fields and a date field. A new 'number(1) default
> 0'
> > is beeing added
> > Assuming the update will run as the only user connected what could be
done
> > to speed up things?
> > Will any of the following help: Disabling triggers? Disabling
constraints?
> > Dropping indexes and rebuilding?
> >
> > Also how do I calculate the how much rollback space is needed?
> >
> > TIA
> > Peter Laursen
> >
> >
> >
> >
>
>
Received on Thu Aug 22 2002 - 12:44:14 CDT
