ORA-00600: internal error code, arguments: [25027], [3], [0], [], [], [], [], []

From: Fuad Habash <fmhabash_at_gmail.com>
Date: Thu, 8 May 2008 14:17:10 -0400
Message-ID: <db523fc50805081117l74865bc9lcd6bc7640cfde35d@mail.gmail.com>

Just wondering if any of the DBA have seen this issue before.

Oracle returning
ORA-00600: internal error code, arguments: [25027], [3], [0], [], [], [], [], []

We narrowed it down to an insert statement in a message_store table which looks like this
SQL> desc tracker02.message_store

 Name                                      Null?    Type
 ----------------------------------------- --------
 MESSAGE_ID                                NOT NULL VARCHAR2(100)
 SENDER_MESSAGE_SIZE                                NUMBER
 SENDER_SMALL_DOCUMENT                              VARCHAR2(4000)
 SENDER_DOCUMENT_DATA                               BLOB
 RECEIVER_MESSAGE_SIZE                              NUMBER
 RECEIVER_SMALL_DOCUMENT                            VARCHAR2(4000)
 RECEIVER_DOCUMENT_DATA                             BLOB
 SENDER_DOCUMENT_DATA_ON_WIRE                       BLOB
 COMMON_MESSAGE_SIZE                                NUMBER
 COMMON_DOCUMENT_DATA                               BLOB
 COMMON_SMALL_DOCUMENT                              VARCHAR2(4000)

To be more specific, the error returns when inserting into a blob column.

These are the facts

  1. Oracle suggested "This looks like Bug 4899479 which causes memory corruptions due to the use of in-memory_undo which is a default in".
    • Per their suggestion, we disabled it and applied the one-off patch. It did not work.
  2. we bounced the instance afterwards, still did not work. Any insert into this table will abort with above error.
  3. The persistence of this error after the reboot and the patch, made us suspect logical/physical corruption.
    - dbv: returned no errors
    • RMAN validate: no errors
    • validate structure: returned ORA-01499: table/index cross reference failure - see trace file
    • the trace file returned row mismatch in index tsn: 5 rdba: 0x11c0098d when we followed the rdba hint, we isolated it to a the PK index on the table. When we rebuilt it, it errored out with duplicate value error. Eventually, we did a disable/enable on the PK and dumped dups into exceptions table. This process returned to rows with same message_id PK!!! Once we deleted one of the rows, the index rebuilt successfully. However, the 600 error persisted.
    • As a last resort, we truncated the table which resolved the issue.

 any comments or suggestions are welcome.

Thank you

Received on Thu May 08 2008 - 13:17:10 CDT

Original text of this message