Re: import using imp 10.2.0.2 commits every row

jasper wrote:
> i try to import a dump created by exp 10.2.0.1 with 10.2.0.2.
> import is running very slow (approx. 1100 rows with avglen=150 bytes
> per
> second). the importing sessions shows a lot of log file sync wait
> events which
> is strange because buffer is set to 2M. as a test i imported only one
> table
> with 140 rows from the large dump file and checked the user commits for
> the
> importing session: 145 commits.
> i repeated the test with different tables with different number of
> rows, user
> commits was every time equal to number of rows + 5. so it seems that
> imp
> commits every row. if i import with commit=n i get 6 commits for
> 140-rows table
> which is plausible. i can reproduce the problem on AIX and linux.

Mayhaps:
http://download-west.oracle.com/docs/cd/B19306_01/server.102/b14215/exp_imp.htm#sthref2398

"For tables containing LOBs, LONG, BFILE, REF, ROWID, or UROWID columns, array inserts are not done. If COMMIT=y, Import commits these tables after each row."

>
> here are the commands:
>
> $ exp userid=user file=dumpfile statistics=none owner=owner direct=n
> $ imp userid=user file=dumpfile commit=y fromuser=owner
> touser=new_owner statistics=none tables=table buffer=2000000
>
> TAR is opened but i would like to get a confirmation if possible.
> anyone?
>

jg

--
@home.com is bogus.
Revolting linux http://www.internetnews.com/dev-news/article.php/3614721