Re: Using oracle dump (quickly!)

"Ron" <support_at_dbainfopower.com> wrote in message news:<vfednQwwoZgWbqfd4p2dnA_at_comcast.com>...
> Hello Dan,
>
> One thing: check imp parameters
>
> - looks like imp would run with commit=n - verify with vendor if
> there is reason for it. If you'd like to stay with commit=n - make sure your
> rollback segments are big (not just rollback tablespace)
>
> - set BUFFER parameter to high number (50M+- depending on a free
> memory)
>
> - set instance LOG_BUFFER parameter to the big value (4M+) for the time
> of import
>
> - see if you can alter tables to "nologging" for the time of import
> (switch them back to logging after)
>
> To Group: please feel free to add.
>
> Regards,
>
> Ron
> DBA Infopower
> http://www.dbainfopower.com
> Standard disclaimer:
> http://www.dbainfopower.com/dbaip_advice_disclaimer.html
>

1 Usually a buffer parameter of 8M is more than sufficient 2 This is just unprofessional advice. Setting LOG_BUFFER too anything higher than 1M is useless, as the buffer is written to disk when 1M is dirty.
3 Nologging doesn't have any impact on import at all. It is only applicable using direct path operations, ie in inserts using the /*+ append */ and using sqlloader in direct mode. Import doesn't work in direct mode.

Could you *please* verify your advice against the docs, before posting it?

The OP would better leave out creating indexes (indexes=N) then run imp with indexfile=<any filename> edit this file changing the logging attribute of the create index statement in nologging , and run this file against the database.

Sybrand Bakker
Senior Oracle DBA Received on Tue Feb 24 2004 - 06:47:11 CST