Re: SQL*NET connections

From: Sybrand Bakker <postbus_at_sybrandb.demon.nl>
Date: Sun, 2 May 1999 13:39:53 +0200
Message-ID: <925645054.5647.0.rover.d4ee154e_at_news.demon.nl>


The problem here is as far as the server side is concerned a client can disappear in a number of ways, including just shutting the local pc. Eventually PMON on the server will discover the session is defunct and kill it.
This can be resolved in two ways:
set sqlnet.expire_time in sqlnet.ora on clients and server to a non zero value, say 10. Now sqlnet will poll every 10 secs to see whether there is still connectivity with the client. This of course causes additional overhead.
The other solution is to configure the multithread server. In this case one or more server processes will communicate with multiple clients. This is based on the assumption from the perspective of the database the average client is doing 'nothing'. Check out your documentation or technet.oracle.com on how to configure MTS.

Hth,
Sybrand Bakker, Oracle DBA

Gopi Balasingam wrote in message <372B534F.F816B0CF_at_solect.com>...
>Hi All,
>
>Sorry for the cross post !
>
>I have been running into some strange problems where the SQLNET
>connections don't seem to terminate once the session is completed. I
>have been testing some apps written for the Windows which use the
>ODBC-SQL*NET layer to connect to the Oracle database instant, but upon
>doing so any subsequent connections are spawned off as another process
>on the UNIX box running the database! Is this a limitation to the way
>SQLNET functions or do all session need a dbclose method in order to
>terminate the session. If so, this would cause a major problems when
>running WEBAPPS that use the SQLNET layer because of the simultaneous
>high hit/connection rate to the database.
>
>Any advice as how to over come this limitation ?
>
>Thanx for you help.
>
>Regards,
>
>Gopi
>
Received on Sun May 02 1999 - 13:39:53 CEST

Original text of this message