Re: What can be the cause for corrupted index?

andreyNSPAM_at_bookexchange.net (NetComrade) wrote in message news:<40ca1b1d.360187422_at_localhost>...
> Today I've got a
> ORA-00600: internal error code, arguments: [12700] , [34201],
> [172070506], [92], [], [], [], []
>
> After going to metalink, and reading
> http://metalink.oracle.com/metalink/plsql/showdoc?db=NOT&id=153788.1&blackframe=0
> and
> http://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_id=155933.1&p_database_id=NOT
>
> tools/queries provided have helped me determine that I had a corrupted
> index (index was pointing to a rowid that didn't exist)
>
> However, data corruption has never happenned to me before, besides
> cases where serious hardware problems existed.
>
> Do these corruptions usually happen for a reason?
>
> The only thing I can think of, 3 days ago we had a situation where 2
> servers were attempting to start the database at the same time (it's a
> Veritas Cluster environment, both servers managed to mount the same
> file systems, one server went into 'sleep' mode, the other started the
> db, then first 'woke up'), which resulted in an ORA-1122 at one of the
> oracle starts (scared me for a while). However, the second server
> never got to 'open' the database, just mounted.
>
> I guess I should run dbv on all datafiles now just to make sure.
> Theory goes that there are mechanisms in place to prevent writing from
> 2 machines to the same file.. I guess I should go dbv all files now
> just to make sure.
> .......
> We use Oracle 8.1.7.4 on Solaris 2.7 boxes
> remove NSPAM to email

I've had 4 cases of database corruption in the last three years 3 of them happened on VERITAS file system. The more layers of software between ORACLE and the file system the more chances for corruption.

I am sure there will be DBAs who ran years of trouble free 24x7 production databases on VERITAS but it wasn't the case for me and you are another example of such misfortune.

Good thing it wasn't a table but then again it might show up later on - keep an eye on your alert log for "Corrupt block relative dba" messages.

In addition to DBVERIFY I'd also run "validate structure cascade" plus FULL EXP just to rule out things that DBVERIFY can't handle.

-VM Received on Fri Jun 11 2004 - 23:58:10 CDT