Errors ORA 03113 and ORA 07445

Sorry for such a long post, but here's the scoop:

We have Oracle 8.0.5 running on a Sun Box -- 2.7. We have four database instances using this box, with 3.5 GB allocated as shared memory, with the four using just over 3 GB. Just over the last couple of days, Oracle seems to be core dumping for a couple of reasons on each database, with the client user receiving the error message of ORA - 03113, which is just a generic end-of-file on communication channel error message.
In the Alert Log we receive the error message of ORA 07455, another "generic" error message, basically meaning that an OS exception occured generating a core file, which does happen. There have been two occassions where one of the instances core dumped: 1. When a client connection/user does a describe on a particular table. This only happens on some clients thru SQL Plus, and not others.

        The message in the alert log is the following:
            ORA-07445:  exception encountered:  core dump [FF2506BC]

            What I find odd here is the "Address not mapped" but maybe this is generic with no meaning?

2. The most recent happened tonight, when a developer was attempting to modify a user created package with a create or replace statement.

    The message in the alert log is the following

        ORA-07445: exception encountered: core dump
[kauGetRowid()+108] [SIGSEGV] [Invalid permissions for mapped object]
[2528164] [] []

        Again, what I find odd is the "Invalid permissions for mapped object"

The problems are not client operating system dependent...I have been able to get the core dumps to happen on both Unix and NT Boxes. So far, the only occurences have been with the describe action, and the package replacement action. However, it doesn't happen for everyone...the developer replacing the package can describe any table and not core dump, yet core dumps when trying to replace the package. This behavior has not been seen through applications using the database, and does not happen with a direct Svrmgrl or SqlPlus connection to the server with the databases.

A couple of tickets have been opened and Oracle has been working on the problem (they have the trace files) and the responses range from installing the 8.0.5.2 patch, to rebooting the client machines to the databases not being clean, so reboot all of them to clear the semaphores. However, these are not new instances, so why this behavior all of the sudden? Is there some kind of linking problem; network problem?????????

A couple of notes: The machine has moved to a new location, and is on a different network. But if it were suddenly a network problem it seems the applications would be choking, and core dumps would happen for more than just desc and replacing a package (doesn't happen for select, insert, update, etc.).

Again, sorry for the long post, but I've been pulling my hair out, and this has to get fixed.

Any ideas?

Thanks in advance.

.. Received on Tue Sep 14 1999 - 22:52:49 CDT