Re: Disappearing Sessions - Revisited
Date: Wed, 20 Feb 2008 00:46:01 -0800 (PST)
On Feb 20, 1:32 am, "Dereck L. Dietz" <diet..._at_ameritech.net> wrote:
> This is in response to the matter of sessions just disappearing without a
> trace. Does this statement make sense?
> "If you are using sqlplusw, some times the session lockup without giving
> control back to the foreground of the application."
are you absolutely sure there are completely no traces left after the session "disappears"? No trace files whatsoever, no core dumps, no entries in the alert log or OS logs? For any abnormal termination there must be some trace left somewhere. I've never seen a case when Oracle process dies absolutely silently unless it successfully completes and disconnects, though I must admit that I certainly didn't have a chance to see all and every bizarre Oracle bug in action. :) Maybe this is the case? That your process hits some obscure exit condition and successfully terminates way before it is supposed to? Maybe you should try instrumenting the code a bit with dbms_system.ksdwrt() and ksdfls() calls before or after each atomic operation in your process so that no matter how the process dies, there will be a trace of its execution progress up until the point where it stopped and you will at least be able to see at which stage it died and approximate when. Then you will be able to investigate other Oracle and system logs at around that time for evidence of abnormal termination and its reason, or inspect the code and figure out what caused it to exit prematurely.
Vladimir M. Zakharychev
N-Networks, makers of Dynamic PSP(tm) http://www.dynamicpsp.com Received on Wed Feb 20 2008 - 02:46:01 CST