Troubleshooting Internal Errors

Abstract of Tamzin Oscroft wrote at Oracle Magazine

Thu Jan 20 13:35:52 2011
Errors in file /DATA/oracle/admin/
prod/udump/prod_ora_2131.trc:
ORA-00600: internal error code,
arguments: [ktfbtgex-7], [1015817],
[1024], [1015816], [], [], [], []

ORA-600 or ORA-7445: What Is the Difference?

ORA-600 is a catchall message that indicates an error internal to the database code. The key point to note  is that it is signaled when a code check fails within the database. At points throughout the code, Oracle Database performs checks to confirm that the information being used in internal processing is healthy, that the variables being used are within a valid range, that changes are being made to a consistent structure, and that a change won’t put a structure into an unstable state. If a check fails, Oracle Database signals an ORA-600 error and, if necessary, terminates the operation to protect the health of the database.

The first argument error message indicates the location in the code where the check is performed; in the example above, that is ktfbtgex-7 (which indicates that the error occurred at a particular point during tablespace handling).

An ORA-7445 error, on the other hand, traps a notification the operating system has sent to a process and returns that notification to the user. Unlike the ORA-600 error, the ORA-7445 error is an unexpected failure rather than a handled failure.

ORA-07445: exception encountered:
core dump [kocgor()+96] [SIGSEGV]
[ADDR:0xF000000104] [PC:0x861B7EC]
[Address not mapped to object] []

 
the failing function is kocgor, which is associated with the handling of user-defined objects. The trapped signal is SIGSEGV (signal 11, segmentation violation), which is an attempt to write to an illegal area of memory. Another common signal is SIGBUS (signal 10, bus error), and there are other signals that occur less frequently, with causes that range from invalid pointers to insufficient OS resources.

Both ORA-600 and ORA-7445 errors will 

• Write the error message to the alert.log, along with details about the location of a trace containing further information
• Write detailed information to a trace file in the location indicated in the alert.log
• In Oracle Database 11g Release 1 onward, create an incident and place the relevant files in the incident directory in the location defined by the diagnostic_dest initialization file parameter
• Write the error message to the user interface if the server process is not terminated or signal ORA-3113 if it is

Often you will see multiple errors reported within the space of a few minutes, typically starting with an ORA-600. It is usually, but not always, the case that the first is the significant error and the others are side effects.

Can I Resolve These Errors Myself?

Check the whole Oracle Magazine article

Help plan the future of Oracle DB Training and Certification

Are you an Oracle Database Administrator?

Would you like to help define the depth and scope of future Oracle Database training and certification?

Oracle is conducting this online Job Task survey prior to the next major Oracle Database product release to ensure that the tasks Oracle Database Administrators consider important are prominent in the design of future Database Administrator curriculum and certification offerings.

Join with other Oracle experts to take this online survey and tell us what tasks are important to you.

Please click on the link below to get started:
Oracle DBA Job-Task Analysis Survey

After completing the survey, get your FREE Oracle Database 11g Interactive Reference Guide. This free interactive reference guide can help you write custom data dictionary scripts, locate views pertinent to a specific database component, and more.

If you have problems accessing the survey, please contact Oracle at DBA-Survey-help

Simple way to get CPU usage

Often we want just the CPU usage from a query rather than complete 10046,

As Tom Kyte published in setember Oracle Magazine:

”

SQL> begin
   2   :cpu := dbms_utility.get_cpu_time;
   3   dbms_application_info.set_client_info(0);
   4 end;
   5 /
PL/SQL procedure successfully completed.

SQL> select owner, f(owner) from stage;
…
72841 rows selected.

SQL> select
  2  dbms_utility.get_cpu_time-:cpu cpu_hsecs,
  3  userenv(‘client_info’)
  4  from dual;

CPU_HSECS   USERENV(‘CLIENT_INFO’)
———   ———————–
       118                     72841
”

To read the whole article:

http://www.oracle.com/technetwork/issue-archive/2011/11-sep/o51asktom-453438.html

Ucertify evaluation

Once again Ucertify gave me the honor to evaluate their products.

This time the product was PMI RMP – PMI Risk Management Professional Prepkit.

The overall feeling is that Ucertify have a very good solid product.

The specific area evaluated areas were:

Download , Installation and registration: Perfect.

Usability: Excellent.

Methodology: Very good.

Quality of questions: Very good.

Answer explanations: Good.

Rough idea regarding database usage

Often we need answer these questions:

How  balanced are the usage among my RAC instances?

What is the database with higher resource demand? (not just redo writting).

Well the query below intend to answer those questions. It shows how much is the usage of database per second since the database started.

select
t1.inst_id,t1.STARTUP_TIME,trunc(value/((sysdate-STARTUP_TIME)*24*60*60)/1000000,2)
from
gv$instance t1,gv$sys_time_model t2 where
t1.INST_ID=t2.INST_ID and t2.stat_name=’DB time’
order by 2;

Futher improvements:

One may create a table , to record snapshots (AWR already do it, but you need the license), then you can check the delta between couple of minutes or hours.