Re: Simple insert - 36,000 physical reads

If you grab the SQL and run it yourself in a session manually (using autotrace or sql trace) do you see the same effect ?

For things like this a SQL trace would also reveal where the IO takes place (file#, block#) and any recursive sql occurring through triggers, vpd, or whatever else there might be.

Stefan

Stefan P Knecht
Senior Consultant
Systems Engineering

OPITZ CONSULTING Schweiz GmbH
Seestrasse 97
CH-8800 Thalwil

Mobile +41-79-571 36 27
stefan.knecht_at_opitz-consulting.ch
http://www.opitz-consulting.ch

OCP 9i/10g SCSA SCNA

On Fri, Dec 12, 2008 at 9:55 PM, Dennis Williams < oracledba.williams_at_gmail.com> wrote:

> List,
>
> Oracle 10.2.0.4, Solaris 8
>
> I've found a simple insert that has:
> 1 execution
> 2,512,102 Buffer Gets
> 36,479 Physical Reads
> 1 Parse call
> 1 Row
>
> This insert statement appears quite a few times in the AWR report, and for
> the times I see it, the statistics are pretty close to what I've listed. The
> insert just has a list of columns and a list of values. No subqueries. There
> is a TO_DATE conversion on two columns.This is a real table, not a view.
>
> Does anyone have any suggestions of why a simple insert could cause this
> much database activity?
>
> Thanks to Brandon for suggesting awrsqrpt.sql in response to another
> question. That is a great tool.
>
> Dennis Williams
>

--
http://www.freelists.org/webpage/oracle-l