RE: Help with Shared Pool Problem

The version of Oracle is 8.1.7, but I had a similar problem on 7.3.4. In both cases the application is doing a mass insert of data, probably without bind variables. For recoverability and restart ability a commit is only issued at the end of the process. From what you have all said, this is causing the shared_pool to fill up with thousands of similar but not identical SQL insert statements. This is causing the problem

Thanks for your help!
Ron

-----Original Message-----
From: Bobak, Mark [mailto:Mark.Bobak_at_il.proquest.com]=20 Sent: Thursday, October 14, 2004 1:43 PM To: Paul Drake
Cc: ganesh.raja_at_gmail.com; Smith, Ron L.; shaharul.anuar_at_intecbilling.com; oracle-l_at_freelists.org Subject: RE: Help with Shared Pool Problem

Good point, Paul.

See this item over at AskTom for full details: http://asktom.oracle.com/pls/ask/f?p=3D4950:8:5150260736162217859::NO::F4= 9
50_P8_DISPLAYID,F4950_P8_CRITERIA:17989406187750, Original poster: What version of Oracle are you using???

-----Original Message-----
From: Paul Drake [mailto:bdbafh_at_gmail.com] Sent: Thursday, October 14, 2004 2:27 PM To: Bobak, Mark
Cc: ganesh.raja_at_gmail.com; rlsmith_at_kmg.com; shaharul.anuar_at_intecbilling.com; oracle-l_at_freelists.org Subject: Re: Help with Shared Pool Problem

IIRC, in 9.2.0.5 - the behavior of open_cursors changes. I remember while attending a presentation at NYOUG (http://www.nyoug.org) given by Steven Feuerstein, that if your code is keeping cursors open that are not using bind variables, that you are in serious trouble.

Wait - it might have been at Jonathan Lewis' seminar. I'll have to get back to you on that.

Check as far as the number of session_cached_cursors and open_cursors. Perhaps Jonathan might have something to add.

Paul

On Thu, 14 Oct 2004 11:55:52 -0400, Bobak, Mark <mark.bobak_at_il.proquest.com> wrote:
> Ganesh,
>=20
> While ORA-4031 can mean that there is no space in the shared pool, it=20
> can also mean there is not enough contiguous memory in the shared=20
> pool. So, if you need 1k for a particular SQL statement, and the=20
> largest available chunk is 900k, then=3D20 Oracle will signal an=20
> ORA-4031. In this case, sometimes, flushing the shared pool can help,

> but not always. =3D20
>=20
> Yes, there is an LRU mechanism for certain components in
> the shared pool. See the scripts I previously mentioned
> on Steve Adams' website, IxOra.
>=20
> -Mark
>

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