SQL*Net: Keepalive

From: Tony Jambu <aaj_at_cmutual.com.au>
Date: Thu, 15 Apr 1993 22:49:41 GMT
Message-ID: <1993Apr15.224941.4482_at_cmutual.com.au>


UNIX/SQL*Net Gurus,

I would appreciate any help on this keepalive problem we are having on site.

Background Information:


  1. Software: Oracle RDBMS: V6.0.33 & V6.0.36 SQL*Net: 1.2.7.2.4 TCP flavour: Wollongong
  2. Hardware: UNIX Servers: SUNs, Solbournes & Pyramids Clients: PCs and X-terminals

Situation:


We regularly have the situation where a PC that is running an Oracle program (eg SQL*Forms) talking to a remote database on a UNIX server via SQL*Net, hangs up and needs to be rebooted.

This will fix the PC's problems but it causes problems at the server end where there is still a connection open and running. It eventually creates all sort of locking and performance problems.

While SQL*Net implements 'keepalive' as a default, it really relies on the Operating/sys (UNIX) to implement it. Does anybody want to guess the default time out interval before there is a time out? 2 hours!!

This is obviously too long. I know it is possible to reduce this and have experimented with it by setting it to 5 minutes and it was successful.

There is now an argument between the DBA (me) and the System Administrator about the time time-out interval to use. His argument is that it will create a lot of network traffic and may cause other programs to die. I would like to have the time-out interval to be between 1 - 2 minutes and the system Administrator say 1 hour. As 5 minutes was unacceptable to both of us, we decided not to change it.

HELP/QUESTIONS:


  1. Has any site changed the keepalive interval and what to?
  2. What processes and programs will be affected eg 'telnetd' & 'rlogind'
  3. Do you know what are the side effects?
  4. What do you recommend?

*5. Will SQL*Net V1 and/or V2 implement their own implementation of

    'keepalive' and not rely on UNIX.

6. What does the following sections in the tcp/doc/README.doc mean

  Version 1.2.7.1.2

  1. Keepalives enabled by default This prevents problems which can occur when a client is disconnected unexpectedly, for instance, if a PC is powered down. The associated shadow process will exit cleanly.

    and    

t:host[/service]:SID[,buffer-size][:conn-retries,keepalive,VSN1,break-mode]
						      ^^^^^^^^^



Any information will be of great help.

Regards
Tony

-- 
 _____       ________ / ____ |Tony Jambu, Database Administrator
  /_  __       /_ __ /       |Colonial Mutual Invest Mgmt,Aust (ACN 004021809)
 /(_)/ ((_/ \_/(///(/_)/_(   |EMAIL:  TJambu_at_cmutual.com.au
 \_______/                   |PHONE:  +61-3-6418448       FAX:  +61-3-6076198
Received on Fri Apr 16 1993 - 00:49:41 CEST

Original text of this message