Wednesday, June 04, 2008

REQUEST:trace dump in directory cdmp_***

I found cdmp**** folders on Database of every nodes:
They created by Oracle Diag:
(REQUEST:trace dump in directory cdmp***)

$ tail db1_diag_16582.trc

REQUEST:trace dump in directory cdmp_20080604120000
Trace dumping is performing id=[cdmp_20080604120000]....
*** 2008-06-04 12:00:01.546
Trace dumping is done
*** 2008-06-04 12:00:53.232
A dump event msg is rcv'd
REQUEST:trace dump in directory cdmp_20080604120052
Trace dumping is performing id=[cdmp_20080604120052]....
*** 2008-06-04 12:00:54.623
Trace dumping is done

$ls -la
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:49 cdmp_20080604114916
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:50 cdmp_20080604115007
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:50 cdmp_20080604115058
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:51 cdmp_20080604115148
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:52 cdmp_20080604115239
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:53 cdmp_20080604115334
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:54 cdmp_20080604115433
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:55 cdmp_20080604115521
drwxr-x--- 2 oracle oinstall 4096 Jun 4 11:56 cdmp_20080604115610
drwxr-x--- 2 oracle oinstall 4096 Jun 4 12:00 cdmp_20080604120000

They made disk PATH be full.

...........................................

So, I checked about Database.

I found a problem about datafile:

SQL> select * from V$DATABASE_BLOCK_CORRUPTION;

FILE# BLOCK# BLOCKS CORRUPTION_CHANGE# CORRUPTIO
---------- ---------- ---------- ------------------ ---------
49 515632 1 8.3019E+12 LOGICAL
49 463340 5 8.3019E+12 LOGICAL

SQL> select * from v$backup_corruption;

RECID STAMP SET_STAMP SET_COUNT PIECE# FILE# BLOCK# BLOCKS CORRUPTION_CHANGE# MAR CORRUPTIO
---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ------------------ --- ---------
1805 656488622 656488411 28602 1 49 463340 5 8.3019E+12 NO LOGICAL
1806 656488622 656488411 28602 1 49 515632 1 8.3019E+12 NO LOGICAL


After that, I used blockrecover by RMAN:

run{
allocate channel t1 type disk;
allocate channel t2 type 'SBT_TAPE';
BACKUP VALIDATE DATABASE;
BLOCKRECOVER CORRUPTION LIST;
}

..........................
That's resolve; REQUEST:trace dump in directory cdmp cdmp***

.........................

2 comments:

Anonymous said...

you may have had corrupt blocks, but they were not the reason for these trace files being created in general. The trace files are created as standard when using RAC. I hit upon this recently, and they were being created, in my case, due to a known streams bug, but to say that all these trace files were caused only by a corrupt database, which is what you imply, is misleading.

Surachart said...

you right, cdmp may not the reason for block corrupt, but this issue solve by this idea.

Deadlock or some error, we can see cdmp error.