PI47916: ABEND04E RC00E20015 OR RC00E20016 DURING SHUTDOWN 15/08/31 PTF PECHANGE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• ABEND04E RC00E20015 or RC00E20016 during DB2 shutdown.
  The error is preceeced by many RC00C20220 and RC00C20221
  errors.
  ADDITIONAL SYMPTOMS: RC00C90101 DSNKTRAV ERQUAL5012
  DB2 shutdown completes without casting out everything, and
  then encounters down-level data later.
  ABEND0C4 in DSNB5UTX +0050 during DB2 shutdown
  due to a bad MMRB pointer in reg 1.
  MSGDSNR055I DSNR055I  @DBS1 DSNRTIMR RESTART SUSPENDED
  SINCE
  14:26:54.82 IN DSNTLSUS +000004D6 UI21287
  MSGDSNR056I DSNR056I  @DBS1 DSNRTIMR RESTART SUSPENDED ON
  IRLM
  
  REQUEST.  ONE HOLDER OF RESOURCE IS SUBSYSTEM
  ********.  SERVICE
  INFO: IRLM FUNC 02, RESOURCE
  10000021140001DB0135000000000000
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: DB2 data sharing users with UI26969 or       *
  *                 UI26970 applied.                             *
  ****************************************************************
  * PROBLEM DESCRIPTION: Various abends in castout processing    *
  *                      during normal DB2 shutdown, including:  *
  *                        - ABEND04E RC00C20220                 *
  *                        - ABEND04E RC00C20221                 *
  *                        - ABEND0C4 in various DSNB5* csects.  *
  *                                                              *
  *                      Possible storage shortage ABEND04E      *
  *                      with RC00E20015 or RC00E20016, due to   *
  *                      repeated retries of one of the above    *
  *                      abends.                                 *
  *                                                              *
  *                      Possible hang during pageset open or    *
  *                      DB2 restart due to a pageset castout    *
  *                      P-lock being held in SIX mode.          *
  *                                                              *
  *                      Potential lost data updates due to a    *
  *                      table space or index becoming non-GBP-  *
  *                      dependent in spite of incomplete        *
  *                      castout processing during shutdown.     *
  *                                                              *
  *                      Corrupted data can result in any of     *
  *                      the following symptoms:                 *
  *                       - Incorrect output, INCORROUT.         *
  *                       - ABEND04E RC00C90101, RC00C90102,     *
  *                         RC00C90105, or RC00C902xx in         *
  *                         various CSECTs.                      *
  *                       - Data/index inconsistencies reported  *
  *                         by the CHECK INDEX utility.          *
  *                       - Page regression reported by the      *
  *                         DSN1LOGP utility.                    *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  PI38310 (UI26969 and UI26970) modified the pageset P-lock
  downgrade/release logic to immediately fail on a conditional
  request if a castout engine is active for the pageset on this
  member.
  
  During DB2 shutdown, most pagesets get closed by the deferred
  close processor, which the above-mentioned logic considers to
  be "conditional".  So pagesets with active castout are skipped
  by this process.  The continuation of shutdown processing will
  drive a second close for these objects, but they can be skipped
  again due to an incorrect setting of a "P-lock released by
  shutdown" flag.  So shutdown will progress, disconnecting GBPs
  and freeing storage, while the castout engine is still running,
  and the castout engine is likely to abend in various ways.
  
  Data corruption may result if the castout fails to complete and
  the member shutting down was the only updating member.  This is
  due to the pageset losing its GBP-dependency even though it may
  still have modified pages in the GBP.
  
  Outside of DB2 shutdown, a close request which was bypassed may
  leave the pageset castout lock held in state SIX, which will
  delay opens by other members.
  

Problem conclusion

• The pageset P-lock downgrade/release logic has been modified to:
    - Not treat shutdown closes by the deferred close processor
      as "conditional" requests.
    - Ensure that the castout P-lock is restored to state U if it
      was upgraded to state SIX.
    - Not set the "P-lock released by shutdown" indication if the
      P-lock was not in fact released.
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

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

  UI32086 UI32087

Modules/Macros

• DSNB1CPL
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RA10 PSY UI32086

     UP15/10/30 P F510 ®

• RB10 PSY UI32087

     UP15/10/30 P F510 ®

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.