Undo tablespace corruption during RMAN backup

The first time that I tried backing up this database using RMAN, I ran into a corruption issue on the UNDO tablespace. A dbverify detects corruption on the datafile as well. What is strange is that a "backup VALIDATE tablespace UNDO" doesn't indicate any corruption! That is pretty bizzare. Anyone know why?

Short of dropping the tablespace and recreating it (after addressing any objects that were using the undo), any other suggestions?

Thanks.

-------
backup database output

channel ch00: specifying datafile(s) in backupset
input datafile fno=00003 name=/u11/oradata/CRMP/userperson01.dbf
input datafile fno=00027 name=/u11/oradata/CRMP/userfiscal01.dbf
input datafile fno=00060 name=/u09/oradata/CRMP/userCDPd.dbf
input datafile fno=00072 name=/u08/oradata/CRMP/undo03.dbf
input datafile fno=00002 name=/u10/oradata/CRMP/usercodes01.dbf
channel ch00: starting piece 1 at 16-SEP-04
channel ch01: starting incremental level 0 datafile backupset
channel ch01: specifying datafile(s) in backupset
input datafile fno=00005 name=/u13/oradata/CRMP/user01a.dbf
input datafile fno=00028 name=/u05/oradata/CRMP/INDXHEARING01.dbf
input datafile fno=00061 name=/u14/oradata/CRMP/userCDPe.dbf
input datafile fno=00073 name=/u08/oradata/CRMP/undo04.dbf
input datafile fno=00004 name=/u12/oradata/CRMP/tools01.dbf
channel ch01: starting piece 1 at 16-SEP-04
released channel: ch00
released channel: ch01
RMAN-00571: ===========================================================
RMAN-00569: =============== ERROR MESSAGE STACK FOLLOWS ===============
RMAN-00571: ===========================================================
RMAN-03009: failure of backup command on ch00 channel at 09/16/2004 06:33:55
ORA-19566: exceeded limit of 0 corrupt blocks for file /u08/oradata/CRMP/undo03.dbf

-----------------

DBVERIFY output

DBVERIFY: Release 9.2.0.1.0 - Production on Wed Sep 15 11:57:21 2004

Copyright (c) 1982, 2002, Oracle Corporation. All rights reserved.

DBVERIFY - Verification starting : FILE = /u08/oradata/CRMP/undo03.dbf
Page 105402 is marked corrupt
***
Corrupt block relative dba: 0x12019bba (file 72, block 105402)
Bad header found during dbv:
Data in bad block -
type: 2 format: 2 rdba: 0x0012019b
last change scn: 0xf500.ba7f7b81 seq: 0x0 flg: 0x02
consistency value in tail: 0x0981f502
check value in block header: 0x4d7, block checksum disabled
spare1: 0x2, spare2: 0x0, spare3: 0xd200
***
Page 105403 is marked corrupt
***
Corrupt block relative dba: 0x12019bbb (file 72, block 105403)
Bad header found during dbv:
Data in bad block -
type: 2 format: 2 rdba: 0x0012019b
last change scn: 0x2100.bb7f7b82 seq: 0x0 flg: 0x02
consistency value in tail: 0x80822102
check value in block header: 0x480, block checksum disabled
spare1: 0x2, spare2: 0x0, spare3: 0x1600
***
DBVERIFY - Verification complete

Total Pages Examined : 256000
Total Pages Processed (Data) : 0
Total Pages Failing (Data) : 0
Total Pages Processed (Index): 0
Total Pages Failing (Index): 0
Total Pages Processed (Other): 255885
Total Pages Processed (Seg) : 0
Total Pages Failing (Seg) : 0
Total Pages Empty : 112
Total Pages Marked Corrupt : 2
Total Pages Influx : 0
aocdb201:/nbscripts/log>

----------------------------

RMAN - backup VALIDATE output - successful!!

allocated channel: ch01
channel ch01: sid=16 devtype=SBT_TAPE
channel ch01: VERITAS NetBackup for Oracle - Release 4.5GA (00000000)

Starting backup at 15-SEP-04
channel ch00: starting full datafile backupset
channel ch00: specifying datafile(s) in backupset
input datafile fno=00070 name=/u08/oradata/CRMP/undo01.dbf
input datafile fno=00072 name=/u08/oradata/CRMP/undo03.dbf
input datafile fno=00074 name=/u13/oradata/CRMP/undo05.dbf
channel ch01: starting full datafile backupset
channel ch01: specifying datafile(s) in backupset
input datafile fno=00071 name=/u08/oradata/CRMP/undo02.dbf
input datafile fno=00073 name=/u08/oradata/CRMP/undo04.dbf
input datafile fno=00083 name=/u13/oradata/CRMP/undo06.dbf
channel ch00: backup set complete, elapsed time: 00:17:55
channel ch01: backup set complete, elapsed time: 00:17:55
Finished backup at 15-SEP-04

Quote:

What is strange is that a "backup VALIDATE tablespace UNDO" doesn't indicate any corruption! That is pretty bizzare. Anyone know why?

The errors are thrown in the alertlog and v$database_block_corruption. Wonder why they don't simply spit them out to the rman logfile! But that would be too easy wouldn't it.

You seem to have got it..after backup validate check:
In Oracle 8i corruption found with the RMAN validate command is only reported in the alert log. In versions 9i+ you can query the view name v$database_block_corruption to determine what corruption, if any, was found by RMAN. The log files only record the process AFAIK !!

You can do this while RMAN is taking backup.This has to be set in your RMAN backup script.
This will check for corrupt blocks in the datafile specified and if its with the limit set then it would continue with the backup

set maxcorrupt for datafile
212,1,2 to 200;

212,1,2 are datafile nos and 200 is the max corrupt blocks that RMAN will check.If the max corrupt blocks exceeds 200 then it will give the same error as below.

regards
anandkl

Quote:

Originally posted by anandkl
If the max corrupt blocks exceeds 200 then it will give the same error as below.

Yup, I've set max corrupt blocks to get RMAN to continue backing it up.

What's funny is v$database_block_corruption continues to list corruption on datafiles that don't exist anymore! I've dropped a particular tablespace altogether, yet it lists it!