From: brorsg@ibm.net (Guenter Brors)
Subject: Re: Dead Clients with SQL*Net and TCP/IP
Date: 1995/09/09
Message-ID: <42rk84$qoi@news-s01.ny.us.ibm.net>#1/1
references: <41rnmr$2t60@news-s01.ny.us.ibm.net> <dischner-2808951133350001@gkc12a.klch.med.uni-muenchen.de> <423nfq$nln@ujf.ujf-grenoble.fr> <DELLDK.577@shellgate.shell.com>
newsgroups: comp.databases.oracle


ep94bvps@noh71cd.shell.com (Burton V. Peltier) wrote:

>Try upgrading to SqlNet 2.1 with the 'Dead Connection Detection'
>feature. Does exactly what you want.
 
>Also, helps on Unix side if power goes out , reboot , etc.

I have found out that only a combination of the SQL*Net 2.1 'Dead
Connection Detection' and the AIX TCP/IP kernel parameters worked in
all cases. 

For the SQL*Net solution I put the parameter SQLNET.EXPIRE_TIME = 10
(meaning 10 minutes) in the SQLNET.ORA on both the client and server.
It works, but only if the client's TCP/IP is active, that is, the
client must be rebooted and not just powered off.

After changing the TCP/IP kernel parameters with the AIX command 'no
-o tcp_keepidle=1200' and 'no -o tcp_keepintvl=12' even a powered off
client PC was detected. The Oracle shadow process was killed after 10
minutes, whereas the Oracle session disappeared a few minutes later,
probably killed by the SQL*Net parameter after the shadow process was
killed.

Thank you all for the tips I have received.



==================================================
Guenter Brors - Consultant Client/Server Solutions
Internet: brorsg@ibm.net
CompuServe: 100137,2305
Phone: +49(40)6913629   Hamburg
==================================================



