One of the instance failure in Oracle 9.2.0.5 RAC on HPUnix 11i

[Sai Krishna_1]

Hi Friends,

RAC instance is crashing during the transaction runs.
Following errors are obnserved in one of the alert log file .
ORA-07445: exception encountered: core dump [kdxlpu()+2008] [SIGSEGV] [unknown code] [0x8000000100200000] [] []
Sat Apr 30 14:49:25 2005
Errors in file /u01/app/oracle/admin/db9i/udump/db9i1_ora_26438.trc:
ORA-00600: internal error code, arguments: [25012], [15], [0], [], [], [], [], []
Sat Apr 30 14:49:28 2005

For more details please refere attachment .

Please get back to me ASAP.

Please treat this as urgent .

Thanks and Regards,

SAI

[Steven E. Protter]

There should be a core dump file somewhere. You are going to want to analyze it and do a stack trace and let oracle support know what it says.

I've found in the past these types of problems have been solved by putting in patchs and working closely with Oracle support.

ORA-0600 is a very non-specific error code. It means the database picked up a problem but has very little idea what kind of problem it is.

SEP

[Alex Lavrov.]

ORA-600 code is an internal oracle error and there is little you can do about it. Just send to oracle the error message, core dumps if found and they will able to find out what caused it.

[Julio Yamawaki]

Hi,

Your problem is:

Corrupt block relative dba: 0x038110aa (file 14, block 69802)
Fractured block found during preparing block for write
Data in bad block -
type: 6 format: 2 rdba: 0x038110aa
last change scn: 0x0000.00f207e8 seq: 0x1 flg: 0x02
consistency value in tail: 0x139013d4
check value in block header: 0x0, block checksum disabled
spare1: 0x0, spare2: 0x0, spare3: 0x0

You will have to recover this block or always that a query access this block you will have problem.
Also, always put db_block_checksum=true in your init.ora, because this parameter will check the block every time dbwr write to disk.

Regards,

[Julio Yamawaki]

I forgot to tell you: query v$database for file number 14, it´s the datafile with the corrupt block number 69802.

Regards

[Volker Borowski]

Hi,

ORA-7445 can be related to an object that faced a NOLOGGING action. This can not be recovered at all (meaning "recovering" in terms of applying redologs), but needs to be reprocessed by application.
I.e. if this bad block belongs to an Index, which has been created with the nologging option, you need to recreate it.

Do not mix up "recover" and "restore".

Good luck
Volker