Re: More on CPU usage after killing an Oracle session

From: John Fruetel <jfruetel_at_ainet.com>
Date: 1995/06/25
Message-ID: <DAproL.H0s_at_koko.csustan.edu>#1/1


joanne_at_eosdev1.gsfc.nasa.gov (Joanne Woytek) wrote:

>Thanks to all who responded to my post concerning CPU usage being high after
>an Oracle session is killed. For one of the problems I have some more info:
 

>The alter system kill session command marks a session as killed but does
>not remove the session until it can deliver a message to the user that
>the session was killed. The problem is that a typical reason for us to kill
>a session is that the user inadvertantly (or due to system problems) lost their
>window but the Oracle session was not killed. So there is no place for Oracle
>to deliver the message to. The dispatcher process for the killed Oracle session
>apparently continues to try to do something and uses most of a CPU to do it.
>Based on the comment in Chapter 4 of the Administrator's Guide "..the session
>cannot be killed until the operation completes. In this case, the session holds
>all resources until it is killed", my suspicion is that there is nothing to be done,
>but it would be nice if there was a "kill immediate" like the "shutdown
>immediate" which forces the session to be killed no matter what.

I've always just nuked the run-a-way process. Assuming that you're running on Unix, SQL*DBA can tell you the job Unix job number and the system administrator can easily kill the job,

John Fruetel
jfruetel_at_ainet.com Received on Sun Jun 25 1995 - 00:00:00 CEST

Original text of this message