Re: Errors in Trace and Alert Files Concerning Disappearing Processes
Date: Fri, 25 Jan 2008 21:01:42 -0800 (PST)
Message-ID: <50ef222a-2a23-46e3-970f-36a1b0712349@k2g2000hse.googlegroups.com>
On Jan 25, 11:26 pm, "Dereck L. Dietz" <diet..._at_ameritech.net> wrote:
> Out of a total server memory of 16GB, 12 GB is allocated to Oracle. There
> is also 250MB in the PGA. I don't have direct access to the server so I can
> only use what information I can get off of the Enterprise Manager (or
> alert/trace files I get sent to me). Charles Hooper did mention to me that
> Enterprise Manager could give false information about page swapping but,
> that being said, it constantly shows at 100% for Virtual Memory Page
> Swapping.
>
> You mentioned memory leaks. I recall seeing something in a number of the
> trace files I was sent about SGA memory leaks. I'll have to pursue that
> also.
>
> My manager has requested our off-site DBA to put in a SR with Oracle over
> the 7445 error.
>
> Unfortunately we are a datawarehouse which has been configured completely as
> a shared server. We have a lot of long-running processes.
You might take a look at bug 5490054 on Metalink to see if it
applies. The bug is related to Resource Manager and the SWITCH_TIME
setting:
"...But when connecting by MTS, resource manager tries to stop the
session by killing shared serever process. So shared server dies and
ORA-3113 is returned to terminal. Because of dying of shared server
process, PMON tries to re-start shared server process, and next
messages are appeared in alert.log at that timing. . found dead
shared server 'S000', pid = (XX, X)"
There are a lot of bug hits on Metalink for core dump ACCESS_VIOLATION unable_to_trans_pc UNABLE_TO_WRITE - you may be able to find a possible fix, but some of those bug reports have been open for 1.5 years. What is the value of CURSOR_SHARING on the server?
Charles Hooper
IT Manager/Oracle DBA
K&M Machine-Fabricating, Inc.
Received on Fri Jan 25 2008 - 23:01:42 CST