PH19639: GRECP RECOVERY SLOW DUE TO CONTENTION BETWEEN DRSTRT02 AND DRSTRT00

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• GRECP recovery slow issue can be observed when start db
  command is used to recover lots of GRECP objects. One typical
  usage here is: START DB command is used to do GRECP recovery on
  DR site for GBP with AUTOREC set to N.
    With PI75286 applied, the auto-recovery GRECP retry
  threads(014.DRSTRT02) and the start db command
  threads(014.DRSTRT00) could have massive contentions with each
  other. In other words, a lots of timeouts and deadlocks. Hence
  slow down the GRECP recover process.
    Below messages could be found in the MSTR:
  .
  DEADLOCK:
  .
  DSNT375I  -DB2B PLAN=BCT      WITH 444
          CORRELATION-ID=014.DRSTRT00
          CONNECTION-ID=DB2B
          LUW-ID=DSNCAT.SYEC1B.D714FD3F1B1E
          THREAD-INFO=SYSOPR:*:*:*:*:*  :*:*
          IS DEADLOCKED WITH PLAN=         WITH
          CORRELATION-ID=014.DRSTRT02
          CONNECTION-ID=DB2A
          LUW-ID=DSNCAT.SYEC1DB2.D714FCBBB35D
          THREAD-INFO=SYSOPR:*:*:*:*:*  :*:*
          ON MEMBER DB2A
  DSNT501I  -DB2B DSNILMCL RESOURCE UNAVAILABLE 445
             CORRELATION-ID=014.DRSTRT00
             CONNECTION-ID=DB2B
             LUW-ID=DSNCAT.SYEC1B.D714FD3F1B1E=0
             REASON 00C90088
             TYPE 00002007
             NAME TESTTD  .TESTTS4 .00000001
  .
  TIMEOUT:
  .
  DSNT376I  -PB6A PLAN=BCT
           CORRELATION-ID=014.DRSTRT02.
           CONNECTION-ID=PB6A
           LUW-ID=VTAM1.PB6ALU.D709DB7EB573
           THREAD-INFO=SYSOPR:*:*:*:*:*  :*:*
           IS TIMED OUT. ONE HOLDER OF THE RESOURCE IS PLAN=
   WITH
           CORRELATION-ID=014.DRSTRT00
           CONNECTION-ID=PB6A
           LUW-ID=VTAM1.PB6ALU.D709DB2E6A8C
           THREAD-INFO=SYSOPR:*:*:*:*:*  :*:*
           ON MEMBER PB6A
  
   DSNT501I  -PB6A DSNILMCL RESOURCE UNAVAILABLE  328
              CORRELATION-ID=014.DRSTRT02
              CONNECTION-ID=PB6A
              LUW-ID=VTAM1.PB6ALU.D709DB7EB573=0
              REASON 00C900BA
              TYPE 00002001
              NAME TESTDB .TESTTS .00000001
  

Local fix

• BYPASS/CIRCUMVENTION:
  1.)start db2 in MAINT mode to run START DB command.
   In MAINT mode, the 014.DRSTRT02 will not be started.
  2.)use AUTOREC=YES for GBP.
   Db2 will recover most GRECP objects automatically right after
  the Db2 restart complete.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All Db2 for z/OS users.                                      *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * Frequent drain timeouts occur during                         *
  * the DRSTRT02 recovery process for                            *
  * objects with a group bufferpool set as                       *
  * AUTOREC(NO).                                                 *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply corrective PTF when available                          *
  ****************************************************************
  If a manual GRECP recovery is started to recover objects
  marked as such, frequent drain timeouts can occur with auto
  GRECP recovery even if the object's group bufferpool is set to
  AUTOREC(NO).
  The DRSTRT02 routine would check for objects marked GRECP
  and attempt to recover those objects. However, it did not check
  their group bufferpools to make sure that auto recovery was on
  until after it acquired a drain. Once it detected that auto
  recovery was disabled, it would not dedrain before proceeding
  to the next object, possibly causing the drain to be held for a
  long time as drains for other objects are acquired. In this
  scenario, acquiring a drain is a costly process considering
  that no recovery is being done on the object.
  

Problem conclusion

• Logic for the DRSTRT02 routine was modified so that it would
  check if the object's group bufferpool had auto recovery
  enabled before a drain is acquired to proceed with the GRECP
  recovery.
  

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:

  UI68125 UI68126

Modules/Macros

• DSNB1PCK DSNISREC
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RB10 PSY UI68125

     UP20/03/10 P F003

• RC10 PSY UI68126

     UP20/03/10 P F003

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.