Re: Solutions for Dead SQL*Net connections

From: Maxim Volodin <mvolodin_at_tmac.com>
Date: 1995/09/08
Message-ID: <42q2s3$1ts_at_portal.gmu.edu>#1/1


From: Andy Chan <chanandy_at_hk.super.net> Newsgroups: comp.databases.oracle
Subject: Solutions for Dead SQL*Net connections Date: 6 Sep 1995 17:35:13 GMT

Organization: HK Super Net
Mime-Version: 1.0
Content-Type: text/plain

Content-Transfer-Encoding: 7bit
Message-ID: <42km4h$kpm_at_tst.hk.super.net>

Since many of you are interested in my solutions on dead SQL*Net v1 connections, there you are ...

For Solaris 2.3, 2.4

# ndd -set /dev/tcp tcp_close_wait_interval 60000
# ndd -set /dev/tcp tcp_ip_abort_cinterval 60000
# ndd -set /dev/tcp tcp_keepalive_interval 60000

   For permanent effect, you may need to add these lines in startup scripts.

For SunOS 4.x

   Edit 2 kernel files (of course backup the originals first.)

   "/usr/kvm/sys/netinet/tcp_timer.h"

     modify to 
          #define TCPTV_KEEP_IDLE     20
          #define TCPTV_KEEPINTVL     20
          #define TCPTV_KEEPCNT        2

   "/usr/kvm/sys/netinet/in_proto.c"
     add a line
         int   tcp_maxidle = 60;      /* optional */

   Then regenerate the kernel & reboot machine.

For HPUX v9 (on S800)

   Unlike SunOS, HPUX cannot regenerate kernel on NETINET changes. This is due to the fact that HPUX do not have the C sources and headers as in SunOS but only object files in archived libraries. Therefore the easiest way to get around it is to ADB the kernel !

     # adb -w /hp-ux /dev/kmem
       tcp_keepintvl?W 0D150
       tcp_keepintvl/W 0D150
       tcp_keepidle?W 0D14400
       tcp_keepidle/W 0D14400

   If you regenerate a new kernel later on, these steps have to be retaken.

Hope these information are useful to those who are in the dark. If you've comments on them, please discuss.

Rgds. Received on Fri Sep 08 1995 - 00:00:00 CEST

Original text of this message