Re: Oracle SQL*Net & tcp keep alive

From: Jean Anderson <jean_at_gso.saic.com>
Date: 17 Jan 1994 11:26:23 -0800
Message-ID: <2heoov$dhs_at_esosun.css.gov>


In <CJMEun.CAL_at_vtm.be>, gerrit_at_vtm.be writes:

> On our system we have encountered some problems with SQL*Net daemon processes
> on the run !

We have experienced similar problems on Sun 4 if the client dies while trying to connect to the database. As I understand it, orasrv hangs because a listener routes processes to orasrv. It listens to one connect request at a time in exclusive mode. If a process dies before the connection goes through, sql*net doesn't realize the process died; it waits forever doing network reads (and massively accumulating time).

A timeout option lets you tell sql*net to wait for a specified number of seconds rather than forever (see $ORACLE_HOME/tcp/doc/README.doc). I usually set it to 30 seconds like this (orasrv version 1.2.7.3.1):

     tcpctl timeout 30

I also do a

     tcpctl debug

which outputs additional info to the tcp log that helps identify the problem clients.

regards,

 -jean Received on Mon Jan 17 1994 - 20:26:23 CET

Original text of this message