IZ08044: AN INPLACE TABLE REORG SCENARIO CAN CAUSE ABEND AND POSSIBLE CORRUPTION

DB2 Version 9.5 Fix Pack 3b for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 1  for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 2a for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 2  for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 4  for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 3a for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 3 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 4a for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 5 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 6a for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 7 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 8 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 9 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 10 for Linux, UNIX, and Windows

APAR status

• Closed as program error.

Error description

• An inplace table reorg might cause the database to abend if all
  of the following criteria are met:
  - the database has been migrated from v8 to v9
  - there is an overflow pointer (aka tombstone) record that
  points to an overflow that is on the same data page
  - the pointer has a four byte record identifier (v8 format)
  - during the table reorg, the pointer will be converted to the
  v9 format (using six byte record identifiers)
  - the page must be reorganized to reclaim enough free space for
  the updated pointer to be written
  If the above criteria are all met, the page reorg will move the
  overflow record, so that it is not properly updated during the
  inplace reorg.  Subsequent access of this record may result in
  an abend with the following message written to the db2diag.log:
  2007-06-13-08.03.37.945205-240 I672422A521        LEVEL: Error
  PID     : 3190                 TID  : 1           PROC :
  db2reorg (dbname) 0
  INSTANCE: inst                 NODE : 000         DB   : dbname
  APPHDL  : 0-2027               APPID:
  *LOCAL.db2prod1.070613115321
  AUTHID  : DB2ADMIN
  FUNCTION: DB2 UDB, buffer pool services, sqlbLatchPoolRange,
  probe:20
  MESSAGE : ZRC=0x84020009=-2080243703=SQLB_RC_PG_NUM_ERR
            "Page number exceeds max -8100"
            DIA8709E Segmented table page number was too high.
  .
  Running db2dart against the table might show errors similar to
  the following:
  .
        Data inspection phase start. Data obj: 109  In pool: 17
        Error: Attempting to access a page (16777285) which does
  not exist. The total number of pages in this tablespace is
  4160000.
        Error: Unable to read pool page 16777285, from pool 17,
        Error: parent pool 17, object ID 109, object type Data.
        Error: extent map anchor page of object is 9184.
        Error: Failed to determine object relative page number,
  for given pool page 16777285 of object ID 109 pool 17 type (x0)
        Error: Pointer row pointing to invalid pool page 16777285
  for slot 50 (offset x49C)
        Error: in page 24273, pool page 2969905, of Object 109, in
  tablespace 17.
        Error: Page data will be dumped to report.
  .
        Error: DMS Header total free space 1128 not equal to
  calculated value 1124 (8124-116-6884).
        Error: in page 24273, pool page 2969905, of Object 109, in
  tablespace 17.
        Error: Attempting to access a page (16777285) which does
  not exist. The total number of pages in this tablespace is
  4160000.
  .
  There is also the possibility of corruption because the
  incorrect offset is updated on the page.
  

Local fix

• Use offline reorg (not inplace) to reorg any migrated tables
  that may have overflow records (for example, tables with VARCHAR
  columns that are updated such that the lengths increase).  Once
  the tables have been reorged, all pointers will be removed, and
  any new ones created will use six byte record identifiers, so
  this problem can no longer occur.
  

Problem summary

• see problem description
  

Problem conclusion

• First fixed in DB2 UDB Version 9.5, FixPak 1
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

  IZ00501

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  DB2 UDB ESE SOL

• Fixed component ID

  5765F4102

Applicable component levels

• R910 PSY

     UP