PH54116: MQ Z/OS: AN RRS JOB THAT IS CONNECTED TO MQ AND ABENDED IS WAITING IN CSQ3EPX CSECT CSQ3RRSX ROUTINE 23/06/01 PTF PECHANGE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• A hang occurred in MQ for a job using RRS coordination after
  the job abended or was cancelled.
  
  From a dump at the time of the hang, the RRS perpective is:
  CTXCEMGR issued a PC 3506 to RRS to process its end context
  syncpoint exit.  RRS then queued a request to the RRS address
  space to process syncpoint for the associated UR and suspended
  the requestor until syncpoint completes.
  
  There was 1 interest from the MQ Resource Manager. RRS called
  MQ, which had not returned. For the RRS TCB, the linkage stack
  entries show RRS did a PC 81F06, and then there was a PC 30D
  from MQ Load Module CSQ3EPX CSECT CSQ3RRSX routine
  EBACTL_RETRY1. MQ is waiting for ECB DIEWA.lECB.  The RRS TCB
  was looping in EBACTL_RETRY1 waiting for the EBACTL flag to be
  set for a thread.
  
  In the reported case, a cancel or kill of the job was done. The
  related MQ thread was making an MQ API call. Since the work
  being done was from an RRS batch application, the task
  underwent an EB switch. As part of this call an execution unit
  switch (EUS) was required, and the task suspended awaiting the
  request completing. In the reported case, the task was killed
  (abended S422) from USS, and recovery routine EUS1FRRE got
  control. The EUS did not complete in the next 2 seconds, so the
  EBDR flag was set to indicate that recovery should be deferred
  until End of Task (EOT). The recovery routine percolated.
  
  A critical recovery exit in CSQMCPRH was not called which
  would haveve turned on EBACTL in the context ACE EB.
  This later leads to looping in CSQ3RRSX and the address space
  failing to terminate.
  
  The problem only exists when EBDR is set in the context ACE EB,
  which only happens for certain timing windows when a task
  abends (or is cancelled) while waiting for an execution unit
  switch.
  
  This problem is a regression caused by APAR PH38111.
  
  
  Additional symptoms:
  -------------------
  ABEND422 ABENDS422
  
  In the reported case, the RRS application affected was
  Financial Transaction Manager for SWIFT Services for z/OS
  (FTM). The problem can affect other RRS applications.
  
  For the reported FTM case, symptoms included:
  
    DNFF4008E ou-service LT 'ltname': Attempt to initialize
    SFD failed; reason code='Old SFD process found for this LT'.
  
    We issued abort + abort force for the LT (logical terminal),
    but the unix broker task was still running.
  
    We tried to cancel the SFD (SWIFTNet FIN Daemon) broker job,
    but the task was still running.
  
    We then moved the LT's to other LPAR; but the broker job
    on the original LPAR still has the instance.DNF_FSM_SLS.lt
    queue open, so we get this error on other LPAR:
  
    DNFF4400E ou-service LT 'ltname': MQ 'OPEN' operation on
    queue 'instance.DNF_FSM_SLS.lt' of queue manager 'qmgr-name'
    failed; Reason code='2042';
    error text='SLS input queue open error'.
  
    Reason code 2042 means MQRC_OBJECT_IN_USE
  

Local fix

• The problem of the hung RRS job was resolved with an IPL.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of IBM MQ for z/OS Version 9       *
  *                 Release 1 Modification 0 and Version 9       *
  *                 Release 2 Modification 0 and Version 9       *
  *                 Release 3 Modification 0.                    *
  ****************************************************************
  * PROBLEM DESCRIPTION: An RRS application that is connected to *
  *                      MQ hangs after abending or being        *
  *                      canceled.                               *
  ****************************************************************
  After an RRS application that is connected to MQ abends or is
  canceled it may hang.
  
  The hang is caused by a loop that waits for the connected
  application to finish its work inside the Queue Manager's
  Address Space. However the internal state that reflects when the
  connected RRS application is no longer operating inside the
  Queue Manager's Address Space was not being set correctly
  resulting in the hang.
  

Problem conclusion

• The code has been corrected so that the internal state that
  reflects when the connected RRS application is no longer
  operating inside the Queue Manager's Address Space is set
  correctly preventing the connected RRS application from hanging.
  

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:

  UI93081 UI93082 UI93083

Modules/Macros

• CSQ3ID30 CSQCECTX CSQM148M CSQMCLMT CSQMCPRH CSQVEOT1 CSQVEUS2
  CSQVEUS3
  

Fix information

• Fixed component name

  IBM MQ Z/OS V9

• Fixed component ID

  5655MQ900

Applicable component levels

• R100 PSY UI93083

     UP23/08/19 P F308 ­

• R200 PSY UI93082

     UP23/08/19 P F308 ­

• R300 PSY UI93081

     UP23/08/19 P F308 ­

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.