Re: Oracle SQL*Net & tcp keep alive
Date: Fri, 14 Jan 1994 22:34:34 GMT
Message-ID: <CJn5Do.9FD_at_csn.org>
>Yesterday a user crashed his client twice and his daemon process wouldn't
>stop. In fact they started using a lot of CPU resources (almost 95% of the CPU
>was busy for user processes and when I killed them both it dropped to less
>than 3%). A "ps -ef" today around 11:00 show that both processes used more than
>500 minutes of CPU time. The processes started yesterday at 17:53 and 17:57.
>Which means that both processes have kept the CPU busy for almost 100%.
>
>I called Oracle about this problem and they told me it was due to TCP/IP which
>did not signal the SQL*Net that a client had disconnected or that a signal was
>given but it was catched. (=> so what he said was : it's not an Oracle problem
>but a unix problem. We are trying to find a way around this problem with
>sqlnet version 2...)
Don't let Oracle pull your leg on this one. We had *very* similar problem with Squeal-Net, but over SNA LU6.2 connection. If client bombed, his oracle processes would run roughshod over the box. We didn't get to try version 2, as the project was stopped first.
>As a bonus : because of the processes still running -> dbshut didn't work
>properly -> the backup script failed ! (ARGH!!!!!)
We modified the stock dbshut and put it in /usr/local/bin, doing a shutdown immediate to prevent a spinning process from hanging the shutdown, and same with cold backups. (It doesn't hurt a hot backup...aside from resource competition).
Good luck. Keep the pressure on Oracle support for a proper fix.
-- Lynn Osburn AIX and MVS Systems Guy losburn_at_pace.com PACE Membership Warehouse 303-843-8000 x8162Received on Fri Jan 14 1994 - 23:34:34 CET