PK67237: DEADLY EMBRACE WITH QR TCP LOOPING, WAITING ON TASK RUNNING ON RO. THE RO TASK IS WAIT TO RUN ON QR WHICH HAS ONLY 1 PROCESSOR.

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• DFHDSDS3 never does an SVC 1 wait because NEXT_TCP_DISPATCH_TIME
  is several minutes into the past. Then FIRST_EVENT is
  several-minutes-into-the-past time.  NEXT_TCP_DISPATCH_TIME
  has remained stuck at an old time is because the TCP task is
  stuck dispatchable on the RO.
  The 2 key things that come in to play here is that this LPAR is
  a uni-processor, and for some reason the CICS job is running at
  a higher dispatching priority than the DB2 jobs. The reason the\
  RO TCB is stuck in the SVC 1 out of the DB2 module is because
  the DB2 event that needs to run to post the RO TCB can't run.
  And the poster can't run because the DBTRMSTR and DBTRDBM1 jobs
  are at a lower DPH than CICS, CICS is looping, and there is
  only 1 CPU. So the problem is basically a kind of deadly
  embrace.  The QR TCB is looping because the TCP task is stuck
  waiting on the task running on the RO.  And the task running on
  the RO TCB (which has been put into a wait out of a DB2 module)
  is stuck because the QR TCB is looping on the only processor.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All.                                         *
  ****************************************************************
  * PROBLEM DESCRIPTION: CICS region hang at shutdown waiting    *
  *                      for CDBQ to complete.                   *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  A CICS system connected to DB2 is running on a uni-processor. It
  is shutdown and as part of this shutdown the transaction, CDBQ
  is initiated to quiesce the CICS-DB2 attach. The disconnect from
  DB2 requires a terminate identify, which is carried out on the
  RO TCB. The CICS system is running at a higher dispatching
  priority to the DB2 subsystem. The QR TCB is in a loop because
  the TCP task is waiting on the RO TCB, but the RO TCB cannot
  get any CPU on the single processor because the QR TCB is in
  its loop.
  The CICS dispatcher should be putting the QR TCB into a WAIT,
  this would then allow tasks running on other TCBs to get some
  CPU time.
  

Problem conclusion

• DFHDSDS3 has been changed when calculating the first_event time
  to check cstp_waiting and if this is off do not set first_event
  to the value of next_tcp_dispatch_time.
  

Temporary fix

• FIX AVAILABLE BY PTF ONLY
  

Comments

APAR Information

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

  PK64945

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

  UK37491 UK37492

Modules/Macros

•    DFHDSDS3
  

Fix information

• Fixed component name

  CICSTS V3 Z/OS

• Fixed component ID

  5655M1500

Applicable component levels

• R400 PSY UK37491

     UP08/06/25 P F806

• R500 PSY UK37492

     UP08/06/25 P F806

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.