Re: Disappearing Sessions - Can Code Alone Cause It?

From: Dereck L. Dietz <>
Date: Wed, 27 Feb 2008 22:29:01 GMT
Message-ID: <Nulxj.14779$>

"Mark D Powell" <> wrote in message On Feb 27, 5:00 am, "Dereck L. Dietz" <> wrote:
> Oracle
> Windows 2003 Server
> 16GB Memory (93% SGA( 93% Buffer Cache,<3% Shared Pool, <3% Large Pool),
> 7%
> PGA)
> Our off-site DBA is saying our code is causing the disappearing sessions
> because, to paraphrase him: that's the database it has to run in. That's
> why hes telling us to do things like incremental commits.
> My question is: can code alone cause sessions to sometimes run to
> completion but other times just disappear without a trace?
> I have a feeling he's trying to make us do work-arounds rather than trying
> to get to the actual root cause of this problem.

I am not sure I understand the memory percentages you posted but if the PGA is 7% of the 16G and the SGA is 93% then the PGA is probably way too small. Plus some memory should be left for the OS and the posted numbers do not show it.

What kind of database is this (OLTP, Warehouse, DSS, OLAP)?

Are there error messages in the alert log related to the "disappeared" sessions? Are there trace files in the udump destination?

There should be clues. Is connection pooling in use?

HTH -- Mark D Powell --

  • Here is the memory breakdown

Total 16GB of memory on server

Total allocated to Oracle 12,800GB allocated as follows:

SGA - 93.75% (12000MB)

o        shared pool     - 3.5% (416MB)
o        buffer cache    - 94.3% (11312MB)
o        large pool        - 0.4% (48MB)
o        java pool         - 1.3% (160MB)
o        other               - 0.5% (64MB)

PGA - 6.25% (800MB) Received on Wed Feb 27 2008 - 16:29:01 CST

Original text of this message