Re: ORA-8102, why?
Date: Wed, 16 Jan 2008 09:15:48 -0800 (PST)
On Jan 16, 5:20 pm, "shakespeare" <what..._at_xs4all.nl> wrote:
> "Cristian Cudizio" <cristian.cudi..._at_yahoo.it> schreef in berichtnews:71a7ef21-5cb7-4b44-948b-466a5d7d1031_at_s13g2000prd.googlegroups.com...
> > On Jan 16, 3:11 pm, "shakespeare" <what..._at_xs4all.nl> wrote:
> >> "Cristian Cudizio" <cristian.cudi..._at_yahoo.it> schreef in
> >> berichtnews:26ce818d-7d12-4eb7-b0c1-221cf4aae15d_at_x69g2000hsx.googlegroups.com...
> >> > Hello,
> >> > i've an Oracle database with RAC 2 node version 10.1.0.5 on Linux
> >> > Suse SLES9 64 bit. Recently i've
> >> > got an error ORA-8102 during the execution of a procedure that was
> >> > deleting some records. So i have found that
> >> > there are two records (the table has over 90 milions of records) for
> >> > which there are no corresponding index entry on one of the indexes.
> >> > I've read metalink notes 395006.1 and 1081343.6 and i can't understand
> >> > how can it happen. The note 395006.1 say's about "lost write ore some
> >> > similar OS issue" but how can be that there are no database block
> >> > corruptions? The notes say also that dropping and recreating the index
> >> > does not solve the issue. How can it be? is there anyone that can
> >> > explain me that? Since the problem occurs only when i try to change
> >> > the value of the indexed column for the two record or to delete the
> >> > records i've choose to let things as are now, because exporting e re-
> >> > importing the entire table woud be to expensive.
> >> > Thanks,
> >> > Cristian Cudizio
> >> >http://oracledb.wordpress.com
> >> >http://cristiancudizio.wordpress.com
> >> Are you using a function based index on this table?
> >> Shakespeare
> > No, it is a normal b-tree index on one field of type DATE. there are
> > other 5 indexes on other field and they are correct.
> > Thanks,
> > Cristian Cudizio
> I think you misunderstand note 395006.1: it does not say dropping and
> recreating the index won't help. It's a description of the symptom.
> You could drop and recreate the index and see if the error remains.
Yes you are right, i misunderstand note 395006.1, it say that are
"other" indexes corrupted
with same problem.
But still remains that i found curious that there is no block corruption detected by oracle
but there is a logical corruption that prevents deletion of the records or updating of field
whose values are missing in the index. Those records are there from 02 Jan 2006 and have
never caused problems until we tried to delete them.