The following conditions must be met for the problem to occur:
- MDC table with the COMPRESS YES option.
- MDC table is created in a DMS table space or an automatic
storage tablespace with a small extent size.
- REORG table command with RECLAIM EXTENTS option returns the
empty extents to the table space.
- The table is truncated either explicitly with the TRUNCATE
TABLE command or implicitly with the IMPORT/LOAD command with
the REPLACE option.
The truncate table is called and eventually replayed on the
standby database or by rollforward. When it is replayed, the
dictionary is moved to the first few extents of the table. If
this requires the reallocation of previously freed extents, an
error occurs.
The rollforward might fail with the following error:
SQL1042C An unexpected system error occurred. SQLSTATE=58004
From db2diag.log, the following error occurs:
2013-02-12-09.24.47.342490-300 I39781E2207 LEVEL: Severe
PID : 24626 TID : 46912883386688PROC :
db2sysc
INSTANCE: db2inst1 NODE : 000 DB : SAMPLE
APPHDL : 0-48 APPID:
*LOCAL.db2inst1.130212142447
AUTHID : DB2INST1
EDUID : 152 EDUNAME: db2redom (SAMPLE)
FUNCTION: DB2 UDB, buffer pool services,
sqlbDMScheckObjAlloc,probe:826
MESSAGE : ZRC=0x8402001B=-2080243685=SQLB_EMP_MAP_INFO_NOT_FOUND
"EMP MAP INFO NOT FOUND"
DATA #1 : Object descriptor, PD_TYPE_SQLB_OBJECT_DESC, 88 bytes
Obj: {pool:3;obj:4;type:0} Parent={3;4}
(...)
(...)
DATA #2 : Fix control block, PD_TYPE_SQLB_FIX_CB, 168 bytes
(...)
(...)
DATA #3 : unsigned integer, 4 bytes
1
2013-02-12-09.24.47.342727-300 I41989E505 LEVEL: Severe
PID : 24626 TID : 46912883386688PROC
:db2sysc
INSTANCE: db2inst1 NODE : 000 DB : SAMPLE
APPHDL : 0-48 APPID: *LOCAL.ztoth.130212142447
AUTHID : DB2INST1
EDUID : 152 EDUNAME: db2redom (SAMPLE)
FUNCTION: DB2 UDB, data management, sqldRedoFastTruncTable,
probe:5089
RETCODE : ZRC=0x8402001B=-2080243685=SQLB_EMP_MAP_INFO_NOT_FOUND
"EMP MAP INFO NOT FOUND"
This results in data loss in the problem table. You must
restore the table without rolling forward and extract the data
that you had in the table at the time the backup image was
created. To recover the rest of the table, contact IBM support
to receive a special service option to rollforward all of the
log records, excluding the log records for the problem table.
If this issue occurs on an HADR standby system, you can recover
the table by taking a new backup from the primary machine and
rebuilding it on the standby system.
The fix does not change the rollforward processing, but the
runtime processing by changing the order in which the logs are
written. As a result, transaction log records created using a
older fix pack might receive the problem, even when using the
version of the DB2 software that contains the fix. |
