PH30979: DATA CORRUPTION AFTER REORG COMPLETED WITH RC00 WITH PH28092 INSTALLED, RC00C90206 ERQUAL5007, INCORROUT, ABEND0C4 DSNURBXD

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Multiple issues after PH28092 / UI71385 is applied.
  Reorg utility might end with RC00 but leaving the tablespace in
  inconsistent state.
  There are index related errors on PI during REORG or after
  REORG.
  In particular there are rows with the same primary key while non
  key columns can have different values containing the tablespace
  rids.
  Applications can get abend04E RC00C90206 ERQUAL5007 in DSNIDIFS,
  or can even get incorrect outputs depending on the access path
  chosen .
  Other symptoms include next REORG possibly ending with ABEND04E
  RC00E4030F RC00C90003 in DSNURLOG or CHECK INDEX and REBUILD
  INDEX errors like MSGDSNU713I KEYS MISMATCH / MSGDSNU340I
  ERROR LOADING INDEX DUPLICATE KEY.
  .
  Another symptom is REORG abending with Abend0C4 RC11 DSNURBXD
  When the PI being built has a larger key size than any other
  index being built, the key record is not initialized propertly
  with the correct length value, which can result in the reported
  problems when the key buffer storage is overran due to incorrect
  accumulation of the key record lengths in the buffer.
  This is on a new function code path introduced by PH28092,
  End user can hit this as a regression simply by applying the PTF
  of PH28092 without performing any other enabling action.
  
  
  User can check if the Reorg output contains this message:
  DSNU1160I.... PARTITIONS WILL BE UNLOADED/RELOADED IN
  PARALLEL...
  If the message is there and PH28092/UI71385 is applied then
  the user is exposed to the issue which can cause the
  unavailability of the involved data to the applications,
  as the user needs to fix the bad data.
  In the reported case the client performed the following
  actions to prevent additional duplicate keys:
  - stop db2 applications / prevent all data access to corrupted
    tablespace
  - drop unique index
  - correct the data
  - re-create the unique index
  - start db2 application
  

Local fix

• To avoid the problem from occurring do one of the
  following:
  Run SHRLEVEL REFERENCE instead of change, or disable
  unload/reload partition parallelism via avoiding SORTDEVT
  specification. Other workarounds are using PARALLEL (1), or
  DIAGNOSE TYPE(892) to disable the code introduced by PE Apar
  PH28092.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All Db2 12 for z/OS users with PH28092 /                     *
  * UI71385 applied when REORG TABLESPACE                        *
  * SHRLEVEL CHANGE is run on a table space                      *
  * with a single clustering and partitioning                    *
  * index defined.                                               *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * REORG TABLESPACE SHRLEVEL CHANGE with                        *
  * unload reload partition parallelism                          *
  * encountered an ABENDS0C4 RC11 in                             *
  * DSNURBXD OFFSET0FE8 with PH28092 /                           *
  * UI71385 applied.                                             *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply corrective PTF when available                          *
  ****************************************************************
  With PH28092 / UI71385 applied, a REORG TABLESPACE SHRLEVEL
  CHANGE using data unload/reload partition parallelism on a
  table space with a single clustering and partitioning index
  defined failed with ABEND0C4 RC11 in DSNURBXD + x'0FE8'.
  The reported problem is caused by incorrect program logic where
  storage movements with an incorrect key length is used to fill
  an inmemory buffer.  This results in a storage overlay which
  leads to the reported abend.
  Additional keywords:  DB2OVRLAY/K
  REORG ABENDS04E RC00E4030F in DSNURLOG
  ABEND04E RC00C90206 DSNIDIFS ERQUAL5007 5007
  MSGDSNU713I KEYS MISMATCH or
  MSGDSNU340I ERROR LOADING INDEX DUPLICATE KEY
  during CHECK INDEX
  

Problem conclusion

• Code has been modified to correct the aforementioned scenario
  related to REORG TABLESPACE SHRLEVEL CHANGE execution on a
  table space with a single clustering and partitioning index.
  

Temporary fix

Comments

APAR Information

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

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

  UI72742

Modules/Macros

• DSNURBXD
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RC10 PSY UI72742

     UP20/12/02 P F012 ­

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.