attached Oracle process dies sometimes

[Serafín]

On HPUX 10.1 environment & Oracle 8.0.4:
I have a program that runs once everyday to perfom several SQL queries (about the data of previous day) & exits.
The matter is that the oracle process associated (to my program) dies sometimes, so my program logs:
ORA-03113: end-of-file on communication channel
(We catch all ORA errors & this is the first)

The SQL just before is always:
EXEC SQL SELECT GOLPES_HORA, CANTIDAD
INTO :fpiezash, :icantidad
FROM HIS_REFERENCIAS
WHERE FECHA = ( SELECT MAX( FECHA) FROM HIS_REFERENCIAS
N_REFERENCIA = :REF
AND
FECHA <= TO_DATE( :sfecha2, 'dd-mm-yyyy hh24.mi.ss')
)
AND N_REFERENCIA = :sref;

On file alert_*.log we can several lines like this: ORA-07445: exception encountered: core dump [11] [0] [0] [0] [] []
Today happens & i've launched my program again(with same parameters: date hour minute seconds...) & Oracle don't fall.
Any idea?
Thanks so much

[Steven E. Protter]

Shalom,

HP-UX 10.1? that's pretty old.

To your issue.

This is a problem that is normally solved with tnsnames.ora or listener.ora or sqlnet.ora

There may be a network issue if this configuration goes across a network.

You should use tnsping to confirm network configuration to all items listed in tnsnames.ora and make judicious corrections based on what you find out.

this also could be made worse by the OS version, lack of patching and other issues.

Back up all config files prior to making any changes.

SEP

[Yogeeraj_1]

hi,

I had a look at metalink note Note:164968.1 and this is what they say:

A common situation which might result in this type of error is where a pointer in memory is overwritten in error by a software problem. In this event, when the pointer is used to lookup the data it is pointing at, this may likely reference an area of memory without the process address space.

Another cause of the issue is where a structure holding pointers to code routines (not data this time) is overwritten in the same way. When a pointer is used to call the desired function, the processor will then attempt to execute meaningless instructions, i.e. we will likely try to execute data.

These conditions can occur due to a bug in the Oracle code, or can be caused by external factors such as:

- Application of operating system patches causing referenced libraries to have changed. Therefore after applying any operating system patches, you should always relink the Oracle binaries.

- Application code changes causing invalid/null values to be used.

- Environmental factors such as the wrong Oracle client libraries being used by a client application, due to incorrect settings of environment variables such as PATH, LD_LIBRARY_PATH etc.



Can you please check this? Also, consider applying any relevant patches.

kind regards
yogeeraj

[Yogeeraj_1]

hi again,

if you have metalink access, you further troubleshoot using the tool found with the following note:
Note 153788.1 Troubleshoot an ORA-600 or ORA-7445 Error Using the Error Lookup Tool

hope this helps too!

kind regards
yogeeraj

[Serafín]

I've checked the ora_xxxx.trc file, the variables seems to be Ok. Futhermore i've tried to force to drop Oracle attached process with pointers variables (bad inizializated) on my SQL (input variables & outputs). The only way to get the same effect is to kill the session (instead kill or drop Oracle attached process).
I've splitted the problematic query too...
Database was rebuilt too, & the problem still happend.
Any idea?

[Patti Johnson]

Did you analyze the core file?
Also check virtual memory settings.
please post output from
ulimit -a (for the user starting the DB )

Since the error is not consistent it may be data volume related - too much data and you exceed process memory.

Patti

[Volker Borowski]

Humm,

did you notice this line in the trc file:

File '/dev/async' not present : errno=2

May be you need to check and adjust either your async IO configuration or enable async IO in the kernel ???
Although I am not sure if async IO was already supported with HP-UX 10.20 on Oracle 8.0.4.

Volker

[Serafín]

Hi & Thanks a lot to all

- Patti, here the results of "ulimit -a":
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) 65536
stack(kbytes) 8192
memory(kbytes) unlimited
coredump(blocks) 4290772993
nofiles(descriptors) 2048

- Volker, really i had not realised that "'/dev/async' not present : errno=2" message appears...
the parameter is "fs_async Select Asynchronous Writes" right? Is current value is 0 on our system.

Now I have a theory about a dynamic library made by us (which holds the "problematic" select) than can have something wrong...

[Patti Johnson]

Serafin,

The values for data and stack may be too small. I don't have access to 8.0 documentation, but 10g requires 32K for stack and 1M for data. I usually just set to unlimited.

Patti