Update Stats: Error on RFBLG

[Former Member]

Hello All,

When we update our database statistics using BRConnect, this runs into an error. The error we're getting is given below:

BR0881I Collecting statistics for table SAPR3.RFBLG with method/sample C ...

BR0280I BRCONNECT time stamp: 2009-03-30 14.02.35

BR0301E SQL error -603 at location stats_tab_collect-16

ORA-00603: ORACLE server session terminated by fatal error

BR0886E Checking/collecting statistics failed for table SAPR3.RFBLG

BR0280I BRCONNECT time stamp: 2009-03-30 14.02.35

After this, BRConnect loses it's connection to Oracle and the job fails. Any attempt to create or delete stats for RFBLG fails with the same error. I've seen note 411773 and the setting in DB21 for the table is correct (active flag = R).

I've already updated BRConnect to the latest version (640 patch 50 in our case).

Does anybody know how to overcome this problem?

Kind regards, Wilbert

[Former Member]

Hi,

I am not sure but check the tablespace space too. I think you will get more in alert log anyway.

Hope this helps.

Manoj

[Former Member]

I've tried to create statistics via DB20 for RFBLF alone, but this will lead to the same error. The log of BRConnect also states "BR0871W Pool or cluster table selected to check/collect statistics: SAPR3.RFBLG". The run takes a while to end up in the error though (about 8 minutes)

I checked the Oracle alertlog, and I see these errors:

Errors in file /oracle/NEP/saptrace/usertrace/nep_ora_20628.trc:

ORA-01114: IO error writing block to file 258 (block # 183623)

ORA-27072: skgfdisp: I/O error

Additional information: 183623

ORA-01114: IO error writing block to file 258 (block # 183623)

ORA-27072: skgfdisp: I/O error

Additional information: 183623

Tablespace PSAPCLUD has > 4GB of free space available. I don't see any growth during the statistics run in PSAPCLUD or PSAPUNDO.

[Former Member]

There are a lot of SAP notes regarding consistency of Oracle databases;

perhaps start with 23345 and 540463.

[Former Member]

Wilbert,

You just ID'd your issue - you have a hardware failure.

The statistics are not failing because it's the index for table RFBLG, they are failing because the block is bad on the disk. That block happens to be where this index is located.

You'll need to determine the additional hardware issues here. Maybe an option would be to try to rebuild the index in the hopes that it will get a new block assignment. However, this may not work as Oracle might need to access this bad block to drop the current index. It's worth a try anyway.

The RFGLG index just happens to be the object on this block. It could be any object...

Do you have a resource who can help you resolve the disk issue here? You'd need to correct that really before moving on...

Good Luck,

Tim

[stefan_koehler]

Hello Wilbert,

let's get a little bit closer to this topic.

At first let's clarify some things..

>The statistics are not failing because it's the index for table RFBLG, they are failing because the block is bad on the disk. That block happens to be where this index is located.

@Tim: But you have seen that the problem occurs while writing blocks and not while reading blocks .. so your theory doesn't really fit... also where can you see that the block #183623 from datafile #258 is related to an index? Maybe i am missing something..

The problem

>Errors in file /oracle/NEP/saptrace/usertrace/nep_ora_20628.trc:

>ORA-01114: IO error writing block to file 258 (block # 183623)

>ORA-27072: skgfdisp: I/O error

>Additional information: 183623

Ok so let's focus on this ... could you please check which file is affected and post the output:

shell> sqlplus / as sysdba
SQL> set linesize 99
SQL> SELECT FILE#, TS#, NAME, BLOCKS FROM V$DATAFILE WHERE  FILE# = 258;
SQL> SELECT FILE#, TS#, NAME, BLOCKS FROM V$TEMPFILE WHERE FILE# = 258;
SQL> SELECT TS#, NAME FROM V$TABLESPACE;
SQL> exit
shell> df -m

Regards

Stefan

P.S.: Also the operating system version / release will be a helpful information

[former_member219299]

You´re trying to collect statistics for a Cluster or Pool table this is not recommended. See the following link. SAP Library - CCMS: Oracle

[fidel_vales]

hi, last post was in 2009.

why are you replying?

[Former Member]

Are there any error messages in the Oracle alert log at the time when brconnect fails?

And did you try to collect statistics for this table alone?

regards