PH13496: Q APPLY NEEDS TO CHANGE TO AVOID BOTH BROWSER AND PRUNE THREAD TRYING TO CLEAN DONEMSG SIMULTANEOUSLY.

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• During STOPQ or program stop, Q Apply browser and Prune thread
  may race to clean IBMQREP_DONEMSG entries and attempt to delete
  the same MQ message from RECVQ. If they encounter MQ rollback
  or fail to commit the deleted MQ message then it may cause
  incorrect deletion and consequently a reapply of an already
  applied MQ message on subsequent STARTQ or program restart.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: 1- Q Apply                                   *
  *                 2- SQL Apply users                           *
  *                 3- SQL / Rep users that utilize SQL / Rep    *
  *                 remote bind                                  *
  *                 4- Capture users                             *
  *                 5- Capture users                             *
  ****************************************************************
  * PROBLEM DESCRIPTION: 1- During STOPQ or program stop, Q      *
  *                      Apply browser and Prune thread may      *
  *                      race to clean IBMQREP_DONEMSG entries   *
  *                      and attempt to delete the same MQ       *
  *                      message from RECVQ. If they encounter   *
  *                      MQ rollback or fail to commit the       *
  *                      deleted MQ message then it may cause    *
  *                      incorrect deletion and consequently a   *
  *                      reapply of an already applied MQ        *
  *                      message on subsequent STARTQ or         *
  *                      program restart.                        *
  *                      2- The user cannot identify the table   *
  *                      row being reworked.                     *
  *                      3- SQL / Rep remote bind job modified   *
  *                      to include ASNMONITOR package           *
  *                      4- Capture always checks the Db2        *
  *                      catalog (prior to Db2 v12) to           *
  *                      determine if the table is altered and   *
  *                      needs to be REORG when activating a     *
  *                      subscription.                           *
  *                      5- The results are unpredictable when   *
  *                      starting Capture using the timestamp    *
  *                      option with a non data-sharing system.  *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  1- Races between Q Apply browser thread and Prune thread at
  STOPQ or program start to clean IBMQREP_DONEMSG may lead to
  incorrect deletion from RECVQ. This subsequently leads to
  reapplying of an already applied MQ transaction message with Q
  Apply reporting sql errors like duplicate rows or rows not
  found.
  2- Apply currently neither explains why a rework occurs nor
  provides any details about the row being reworked.
  3- In order to utilize ASNMON remotely the ASNMONITOR package
  must be bound remotely
  4- It is not always possible to determine if REORG is required
  by querying the Db2 catalog. Only the user knows if the change
  was safe.
  5- If you use the timestamp option when starting capture with a
  non data sharing system, the results will be unpredicatble.
  

Problem conclusion

• 1- Q Apply will prevent simultaneous cleanup of RECVQ or
  IBMQREP_DONEMSG by Browser thread and Prune thread at STOPQ or
  program startup by ensuring Prune thread commits and stops
  further pruning of the RECVQ before browser starts the final
  cleanup of IBMQREP_DONEMSG.
  2- Enhance Apply to display the SQLCA and SQLDA when reworking
  an operation if the invocation parameter TRCRWK is specified.
  3- Add the BIND PACKAGE statement for the ASNMONITOR package to
  the ASNBNDDR sample job
  4- Introduce a new Q and SQL Capture startup parameter
  CAPSTART_REORGCHECK=N (default is Y) to turn off checking for
  REORG while capstarting subs.
  5- Change Capture not to start from timestamp if the Db2 system
  is not data sharing. Display error message ASN7268E and stop.
  Note: ASNCLP has been updated from db2_v105fp11 s190723.
  

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:

  UI64865 UI64866 UI64867 UI64868

Modules/Macros

•    ASNAAPP  ASNACMD  ASNACMP  ASNADMSP ASNAFET
  ASNAISO  ASNAMAN  ASNAPPLY ASNAPRS  ASNAWPN  ASNBNDDR ASNCAP
  ASNCATM  ASNCCMD  ASNCDINS ASNCLPAP ASNCLPCL ASNCLPCM ASNCLPCO
  ASNCLPMS ASNCLPQA ASNMCMD  ASNMON   ASNQACMD ASNQAHKT ASNQAPP
  ASNQBRWZ ASNQCAP  ASNQCCMD ASNQDEP  ASNQEXRP ASNQMFMT ASNQXFMT
  ASNRBASE ASNTDIFF ASNTDSP  ASNTRC   ASN2BASE
  

Fix information

• Fixed component name

  WS REPLICATION

• Fixed component ID

  5655L8800

Applicable component levels

• RB33 PSY UI64865

     UP19/08/28 P F908

• RB34 PSY UI64866

     UP19/08/29 P F908

• RB35 PSY UI64867

     UP19/08/28 P F908

• RB36 PSY UI64868

     UP19/08/28 P F908

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.