Re: What application code could cause an ORA-03113 error?

Tom McSally <tom.mcsally_at_xantic.net> schreef in berichtnieuws d0a4b1f.0307160523.64855ca8_at_posting.google.com...
| I have a client who has been getting ORA-03113 errors on most of their
| sites for several months. They have 11 sites using our product, 10 of
| them using our 5.2 version and 1 using our 5.1 version. All of the
| sites using our 5.2 version have been getting ORA-03113 errors, the
| one site using our 5.1 version has NOT been getting the error.
|
| Recently, the site using 5.1 has upgraded to 5.2 and now they too are
| getting the ORA-03113 errors.
|
| This is a pretty clear indication to me that there is something in our
| 5.2 product that is causing these errors but our developers insist it
| is a generic Oracle problem and not a bug in their code.
|
| I need some suggestions as to what our 5.2 product could be doing that
| would cause the ORA-03113 errors. The errors are occurring in our Task
| Server functionality. The Task Server is an application that connects
| to Oracle then sits idle for 1 minute. It then does a query against an
| Oracle table to see if it has a task to perform. This query will
| typically return 0 or 1 row. It continues this 1 minute polling all
| day. Eventually it gets an ORA-03113 error.
|
| I've researched the generic ORA-03113 responses but am more interested
| in what an application could be doing that would cause an ORA-03113
| error after being connected to Oracle for hours.

Oh no, not the ORA-3113 again.
It means that the server process on the database is gone, unexpected. And now you want to know why? Can be many causes, can be very difficult to track.
Once (Oracle V6 was used at the time) I spent - together with Oracle support - finding a cause for 2 years. We suspected it had to do with the 2 clustered tables in the application. We reorganized them to normal tables and have never seen 3113 again..... But a cause is never found.

Check if there are any trace files written on the database server in the user dump dest. Check the alert log in background dump dest for errors too. Nothing there? I was afraid so. So the server process dies so suddenly, it can't even write a goodbye note.....
The "Action" on the doc on the error message suggests to check the Net8 software or network. Well, I suppose nothing changed to the network when they upgrade from 5.1 to 5.2 so very unlikely it's the cause. Then the Net8 software. What versions of Oracle and platform are involved in both 5.1 and 5.2? Is there a change?
Is the 5.2 software that is distributed linked against the proper versions of Net8 and its libraries?

Check if sqlnet.expire_time is set in sqlnet.ora on the client or server machine. If set try to set it to 0.

I suspect the server process sends a net8 message to the client and then an error occurs: the client does not react, the message is invalid or things like that. The server process decides therefore to terminate. Setting Net8 tracing on (on client and/or server) could lead to a cause, but it generates much data.

Good luck. Received on Wed Jul 16 2003 - 09:39:07 CDT