PI21198: WMQ Z/OS LOG ARCHIVE HANGS 15/05/14 PTF PECHANGE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• In the test of adding of new logs to the limit of 31 log
  copies,  Queue manager CSQ1 had 7 dual LOGCOPY, it was added
  from 8 to 32.  On the 33rd log, expected error was displayed:
  CSQJ143I $CSQ1 BSDS active log data set record is full.
  Several manual ARCHIVE LOG commands were entered to ensure new
  logs  were being utilized.  4 new logs were able to be
  successfully archived.  For the rest of the logs, it appears MQ
  writes the logs to tape, but then does not release  the drive
  and catalogue the dataset.
  The situation is this MQ subsystem grabs all our tape
  drives, writes the log and causes the log archive process to
  hang.  Failed to shut down MQ via STOP QMGR command, and
  had to CANCEL it.  After bringing MQ up, it does exactly the
  same thing - allocates the tape drives, writes the logs, but
  does not terminate the log archive.
  *
  Additional Symptom(s) Search Keyword(s):
  If a SET ARCHIVE command is issued at the same time as an
  offload is in progress, the following sequence may occur:
  .
  - The first log archive completes normally.
  .
  - The next log archive fails:
  .
  CSQJ073E LOG ARCHIVE UNIT ALLOCATION FAILED, REASON CODE=0008.
  ALLOCATION OR OFFLOAD OF ARCHIVE LOG DATA SET MAY FAIL
  .
  CSQJ103E CSQJDS01 LOG ALLOCATION ERROR
  DSNAME=<garbage name>
  .
  CSQJ115E OFFLOAD FAILED, COULD NOT ALLOCATE AN ARCHIVE DATA SET
  .
  CSQJ139I LOG OFFLOAD TASK ENDED
  .
  CSQV086E QUEUE MANAGER ABNORMAL TERMINATION REASON=00D10251
  .
  IEA794I SVC DUMP HAS CAPTURED:
  DUMPID=nnn REQUESTED BY JOB (ssidMSTR)
  DUMP TITLE=ssid,ABN=5C6-00D10269,U=SYSOPR  ,C=R3600.710.RLMC-BF
              R-WRT ,M=CSQJW008,LOC=CSQJL002.CSQJW107+0000217E
  .
  <garbage name> in the CSQJ103E message starts with the value of
  the UNIT parameter from CSQ6ARVP with the first letter stripped
  off and  continues with the ARCPFX1 value and other parameters
  from CSQ6ARVP.
  .
  - When the support center reviews the dump, the copy of the
  CSQ6ARVP parameters pointed to by LMBNARVP looks corrupted or
  overlaid.  The control block appears to be shifted right by two
  words, which in the reported case contained zeroes.
  What should start with 005D011A  C1D9E5D7, e.g.
   005D011A  C1D9E5D7  00007000  000010E0  | .)..ARVP.......\ |
  instead is:
   00000000  00000000  005D011A  C1D9E5D7  | .........)..ARVP |
  .
  When UI19842 caused an offset change of 8 bytes to CSQDDPRM,
  CSQJOFF1 was not recompiled to pick up this change.  It
  therefore copies ARVP information from the wrong offset within
  the DPRM to the address pointed to by LMBNARVP.
  .
  A recycle of the queue manager will clear this symptom.
  

Local fix

• N/A
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of WebSphere MQ for z/OS Version 7 *
  *                 Release 1 Modification 0.                    *
  ****************************************************************
  * PROBLEM DESCRIPTION: Queue manager hangs as it archives a    *
  *                      log to tape, but all tape units were    *
  *                      in use. The queue manager does not      *
  *                      recover from the hang when the offload  *
  *                      tasks have completed.                   *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  The offload service task, CSQJOFF1, calls CSQJDS01 to issue SVC
  99 to allocate a dataset on a tape unit. However all tape units
  are in use, so it waits. When CSQJOFF7 offload tasks for
  previous archives complete they post LOAEECB in order to resume
  CSQJOFF1 but CSQJOFF1 cannot resume because it is waiting in
  CSQJDS01, so there is a deadly embrace.
  

Problem conclusion

• A new CSQZPARM parameter, MAXCNOFF, has been introduced to limit
  the number of CSQJOFF7 offload tasks that can be run in
  parallel. This will allow a queue manager or queue managers to
  be tuned such that they will not use all the available tape
  units. Instead the queue manager will wait until a CSQJOFF7
  offload task has completed before trying to allocate any new
  archive datasets.
  100Y
  CMQCFA
  CMQCFC
  CMQCFP
  CSQJC00A
  CSQJOFF1
  CSQJOFF2
  CSQJOFF6
  CSQJS001
  CSQZMSTR
  CSQ6LOGP
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

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

  PI27066 PI28014 UI22931

Modules/Macros

• CMQCFA   CMQCFC   CMQCFP   CSQJC00A CSQJOFF1 CSQJOFF2 CSQJOFF6
  CSQJS001 CSQZMSTR CSQ6LOGP
  

Fix information

• Fixed component name

  WMQ Z/OS V7

• Fixed component ID

  5655R3600

Applicable component levels

• R100 PSY UI22931

     UP14/12/06 P F412 ¢

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.