PH29073: ABEND EC6/AKEB, BPXI039I SYSTEM LIMIT MAXPROCSYS HAS REACHED 100% OF ITS CURRENT CAPACITY CAN ALSO GET 878 AT SHUTDOWN

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• CICS is running a workload that uses the IP CICS Sockets
  interface with OTE=YES.  The first time each task makes a socket
  call a USS process is created and associated with the L8/L9 TCB
  the task is running on.  Later at task termination CICS makes a
  BPX1MPC call to clean up the process associated with the L8/L9
  TCB.  When the next task runs on the L8/L9 TCB and makes a
  socket call a new process is created.  This can cause a
  fluctuation in the number of processes on the LPAR as clean up
  of USS processes is not synchronous.
  BPXI039I messages will be issued if the LPAR approaches the
  MAXPROCSYS limit.  Reports may show a large number of zombie
  processes (processes currently being terminated).  If the number
  of processes reaches the MAXPROCSYS limit then CICS tasks may
  abend EC6/AKEB.
  The abend EC6 will have reason 0F01C008 indicating: C008 The
  calling process cannot be dubbed.
  The following may also be observed:
  JRMaxChild: The maximum number of processes for this user ID has
  been exceeded
  Additional Symptoms:
  You just migrated to CICS TS 5.5.  When shutting
  down the region you receive many ABEND878's.  Here
  are the messages in the log:
  .
  DFHRM0130 applid Recovery manager has successfully quiesced.
  IEA705I ERROR DURING GETMAIN SYS CODE = 80A-10 jobname
  IEA705I 00F86880 0086F788 0086F788 00E51200 00000148
  IEA705I ERROR DURING GETMAIN SYS CODE = 878-10 jobname
  IEA705I 00F86880 0086A588 0086A588 00E61200 000000E8
  .
  A dump from a slip on the 878 abend showed:
  CICS had quiesced when this dump was taken
  Summ Format showed there were a little over 250 TCBs.
  The region was out of storage below the line.
  There was a pattern of storage allocated to these TCBs
  in SP 230 SP230 K0, each having x'E8' bytes allocated.
  Here is what this looked like in VSMDATA 'NOG SUMM'
  .
  DQE:  Addr 0083C000 Size     1000
                      FQE: Addr 0083C000 Size      F18
  .
  System trace showed the failing getmain for x'E8' bytes
  issued out of IEFJRECM+50.
  Browsing these x'E8' storage areas, they have eye catchers
  SSOB  SSIB and MSTRMSTR. Here is an example:
  .
  |.....c.%....SSOB.....c.}.....c..|
  |.........................d.h..=.|
  |.............c..................|
  |................SSIB....MSTRMSTR|
  |    ............................|
  

Local fix

• This problem can be circumvented by applying PH27111 and setting
  the following feature toggle.  This reverts the behaviour of
  CICS to that of CICS 5.4 and earlier where CICS does not clean
  up a process from a L8/L9 TCB at end of task.
  com.ibm.cics.ds.freeussprocesses=false
  Note, in other scenarios the end of task clean up added in CICS
  5.5 is advantageous to ensure that L8/L9 TCBs do not retain an
  associated USS process unnecessarily which can also cause the
  MAXPROCSYS limit to be reached.
  Additional Keywords:  Upgrade2CICS55 Upgrade2CICS56
  subpool scheduler EOT End of Task
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All CICS users.                              *
  ****************************************************************
  * PROBLEM DESCRIPTION: CICS tasks abend EC6/AKEB, BPXI039I     *
  *                      messages indicate that the MAXPROCSYS   *
  *                      limit has been reached.                 *
  ****************************************************************
  CICS is running tasks that use IP CICS Sockets with OTE=YES.
  The first time each task makes a socket call a USS process is
  created and associated with the task's L8/L9 TCB.  At task
  termination CICS makes a BPX1MPC call to release the process.
  The TCB is returned to the pool for use by a subsequent task
  which may cause another USS process to be created.
  
  The release of USS processes is not synchronous.  As a result it
  is possible for the number of terminating processes, also known
  as zombies, combined with the number of active processes to
  fluctuate.
  
  If the total number of processes approaches the LPAR's
  MAXPROCSYS limit BPXI039I messages will be issued.  If the
  MAXPROCSYS limit is reached CICS tasks that attempt to create a
  process will abend EC6/AKEB.  Other work on the LPAR may also
  abend EC6.
  
  Additional keywords: abendsEC6 abendEC6 abendsAKEB abendAKEB
  msgBPXI039I JRMaxChild EZACIC01
  

Problem conclusion

• CICS has been updated so that if the IP CICS Sockets TRUE is
  enabled in a region then CICS will not free USS processes from
  L8/L9 TCBs at end of task.  This allows subsequent tasks to
  reuse existing USS processes, preventing any fluctuation in the
  number of USS processes in existence.
  
  Detection of the IP CICS Sockets TRUE takes precedence over the
  feature toggle added by PH27111.
  
  The Knowledge Centers for CICS Transaction Server for z/OS 5.5
  and 5.6 will be updated as follows.
  
   - The "Upgrading CICS regions" page will be updated to add
     reviewing MAXPROCSYS and MAXPROCUSER as actions applicable to
     upgrades from version 5.1 through 5.4.
  
   - The "The SYS1.PARMLIB(BPXPRMxx) parameters" page will be
     updated to provide more detail on how to calculate the
     MAXPROCSYS capacity required for a CICS region.  An
     explanation of how to determine a suitable MAXPROCUSER value
     will also be added.
  FIXCAT: SWPSP/K
  

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:

  UI71723 UI71725

Modules/Macros

• DFHDSDM  DFHDSIT  DFHDSITT DFHDSKE  DFHUEM
  

Fix information

• Fixed component name

  CICS TS Z/OS V5

• Fixed component ID

  5655Y0400

Applicable component levels

• R200 PSY UI71725

     UP20/09/24 P F009 ¢

• R300 PSY UI71723

     UP20/09/24 P F009 ¢

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.