Re: FRANTIC! ORA-03113 Acessing Specific, Random Tables

"Arielle Smith" <smithari_at_hotmail.com> wrote in message news:677227cc.0209032104.6c0b4b19_at_posting.google.com...
> We are getting ORA-03113 errors when trying to change data in specific
> tables in our Oracle DB and it is never the same table twice. We have
> an application front end that throws an "Oracle connection lost" error
> when the user tries to perform certain actions. One day it will be
> the "data" table, another it will be the "person" table, etc., etc.
> They can perform actions that result in a select statement against it,
> but any simple insert, update, or delete boots them out from any
> client immediately. When I grab the last statement they were
> processing and plug it into SQLPlus on the server, I get the same
> error. They can immediately log back in and perform work that updates
> data in any other table, but consistently get booted every time if
> they touch the underlying table having problems. Only restarting
> Oracle cures it. Unfortunately, it is not automatically logged in any
> trace file or the alert log. We've been rock solid stable for years
> until just the last few weeks. We upgraded to 8.1.7.1.1 on WIN2K
> several months ago without incident. Here are the steps I've
> performed and the results. (Sorry--these steps are all new to me so I
> don't know what's useful and what's overkill.) I found it really
> interesting that the ALTER SESSION listed at the end simply just
> stopped after recording the EXEC statement. One specific question I
> have is what a "shadow process" is. I've seen it referred to in
> multiple postings but am completely in the dark. Also, any ideas on
> other tests I can run to pin it down further? Thank you!:
>
> *****Tested drives and network. All check out clear.

I'd second Billy's suggestion to recheck this We had an Ora-03113 recently that was down to poor network drivers. The network connection functioned just fine apparently, but packet sniffing revealed almost 10% of TCP/IP packets were timing out. Changed the (somewhat exotic) network card configuration and updated the drivers and the problem went away.

In addition as Vladimir says if a process has died on the server side there will be a trace file in the background dump destination on the server. IIRC they are named CORE***.log

> *****Used the client software tracing function (this is where I
> discovered the ora-03113 entry:
> 717393562> SQL HostVar: :hvDEANUMBER =
> 717393562> SQL HostVar: :hvLICNUMBER = (this was the last value in the
> insert stmt)
> 717393859> ERROR: 32786 COULDN'T FIND THE SYMBOL: _MLErrFile (this is
> specific to the app)
> 717393906> ORA-03113: end-of-file on communication channel
> 717393906> ERROR: 32802 ORA-03113: end-of-file on communication
> channel
> ORA-03113: end-of-file on communication channel

Obviously I don't know the app but having one error in a trace file *immediately* before another usually means the first error is the cause. So check this out with the vendor

If you can establish that it really isn't the app and none of our stabs in the dark work out then you should be raising a TAR with Oracle (who may well ask you to patch to 8.1.7.4)

--
Niall Litchfield
Oracle DBA
Audit Commission UK
*****************************************
Please include version and platform
and SQL where applicable
It makes life easier and increases the
likelihood of a good answer
******************************************