Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: internal error

Re: internal error

From: <mike_st0ne_at_my-deja.com>
Date: Tue, 16 Jan 2001 11:52:36 GMT
Message-ID: <941cm2$e68$1@nnrp1.deja.com>

In article <940so7$fig$1_at_SOLAIR2.EUnet.yu>,   "Rexx" <nebacar_at_eunet.yu> wrote:
>
> Hi,
>
> First, sorry for my bad english.
> We have Oracle 7.3.2 on IBM AIX4.2 and have problem:
>
> When I execute SQL statment
>
> SQL> select sif_poslov,trans_seq,br_dok from l_poslovi
> 2 where sif_vr_posla in ( 40,41,42,43)
> 3 and br_dok<0;
>
> no rows selected
>
> but if I execute statment (only difference is br_dok=-1) then I got
 internal
> error
>
> SQL> select sif_poslov,trans_seq,br_dok from l_poslovi
> 2 where sif_vr_posla in ( 40,41,42,43)
> 3 and br_dok=-1 ;
>
> ERROR:
> ORA-00600: internal error code, arguments: [12700], [318891907],
 [27], [],
> [],
> [], [], []
>
> no rows selected
>
> SQL>
>
> I know that support for Oracle 7.3 ended 12/31/2001 but I hope that
 someone
> can help to me.
>
> Thanks!
>
> ------------------------------------------------
> Remove nospam from address
> My e-mail: nebacar_at_eunet.yu
> ------------------------------------------------
>

  I've seen this error at 8i - I would guess it hasn't changed from 7.3, but to confirm it, you would need to call Oracle Support. It's potentially a nasty one.

Basically, this error signals that an index entry points to a missing ROWID. 2 possible causes, one relatively benign, one not:

  1. consistent read problem, caused by a long running query scanning an index which is being updated and index splitting is occurring as a result. If a split block is hit, it's possible that a consistent read back to a pre_split time can have an index entry marked as valid (incorrectly), whereas the corresponding data block has the row flagged as deleted (correctly). (remember that indexes process updates as delete/inserts)
  2. The ROWID really is missing, in which case, the index is corrupt

To find out which is the case, issue the following:

ANALYZE TABLE <table_owner>.L_POSLOVI VALIDATE STRUCTURE CASCADE;

if this throws up an error (probably an ORA-1499 or something similar), you've got a corrupt index, and it will need to be rebuilt - via a "drop" and "create".

   Even if you don't get an error, call Oracle Support (pace Sybrand!), and send them the trace file from the ORA-600 to verify everything is OK.

Hth

Mike S

Sent via Deja.com
http://www.deja.com/ Received on Tue Jan 16 2001 - 05:52:36 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US