PK54949: COLD-START COMMAND /NRE CHKPT 0 ZEROS MESSAGE QUEUE BIT MAPS EVEN IF COMMAND IS LATER REJECTED DUE TO DUPLICATE SSID

APAR status

• Closed as program error.

Error description

• Customer planned to shutdown IMS normally, cold start IMS,
  and reload the message queues using an OEM Tool.
  During shutdown, IMS hung after closing the log, but before
  DBRC signoff (hang was in DFSIDC00 closing VTAM ACB).
  IMS was eventually modified down.
  Later, IMS was brought up, and /NRE CHKPT 0 command entered.
  Because the log had been closed, there were no open log errors
  from DBRC. DFSRST00 makes the call to open system datasets
  before DBRC signon. If cold-start is selected (SCDDATE=0)
  DFSQDOC0 will end up zeroing the message queue dataset bitmaps
  during logical open. DFSRST00 then attempts the DBRC signon,
  fails, issues DFS033I, and waits for another restart command,
  but at this point the message queues have been reset.
  Customer entered /ERE and it was successful. DFSQFIX0 ran
  and found that all the messages queued off all destinations
  were in free DRRNs, and deleted all the messages.
  Message DFS2610 is issued, but otherwise /ERE appears normal.
  However all messages have been discarded. Customer was unaware
  of this and suffered major application impact as a result.
  IMS should not perform any part of cold-start processing
  until after the DBRC Signon check is completed.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All IMS V7 users.                            *
  ****************************************************************
  * PROBLEM DESCRIPTION: Messages were destroyed after an        *
  *                      emergency restart.                      *
  ****************************************************************
  * RECOMMENDATION: INSTALL CORRECTIVE SERVICE FOR APAR/PTF      *
  ****************************************************************
  A hang problem during a controlled IMS shutdown with /CHE DUMPQ
  prevented the DBRC normal signoff from completion.
  
  Subsequently, the attempt of restarting IMS with a cold start
  /NRE CHKPT 0 failed on DBRC signon due to the previously
  failed signoff with DFS033I DUPLICATE ENTRY ON SIGNON REQUEST
  message.
  
  Before having realized the DBRC signon failure, a queue manager
  call to initialize dataset bitmaps was made, which destroyed the
  messages on the queue.
  
  The cold start attempt was aborted and IMS awaited another
  restart command.  The next emergency restart with /ERE completed
  successfully but the messages were lost.
  
  Restart incorrectly clears the dataset bitmaps before verifying
  whether the DBRC signon is successful or not, it should only do
  so after a successful DBRC signon.
  

Problem conclusion

• AIDS: RIDS/SYS RIDS/CNTRL SYS CNTRL
    GEN:
  
  *** END IMS KEYWORDS ***
  The analysis shows that upon the initial DBRC signon failure,
  only the cold start types that lead to restart abort will create
  the problem of the lost messages; such as /NRE CHKPT 0 and /ERE
  COLDSYS, since they allow a subsequent emergency restart to
  complete without IMS initialization. An IMS cold start
  with override flavor, /ERE COLDSYS OVERRIDE, is not subject to
  this problem as it ignores the DBRC signon failure and allows
  IMS restart to continue.
  
  The later DBRC signon failure in restart, at end of log
  processing in DFSRLP00 which leads to the subsequent IMS to
  ABENDU0042, forces a a reinitialization of IMS and the
  message queues to be properly rebuild.
  
  DFSRST00 - RESTART PROCESSOR
  Changed to perform system data set open only after a successful
  initial DBRC signon for /NRE CHKPT 0 and /ERE COLDSYS.
  
  Changed the code portion of system data set open at RESTA021
  label to become a callable internal routine named OPNDSRT.
  

Temporary fix

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

Comments

APAR Information

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

  PK28573

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

  UK31813

Modules/Macros

• DFSRST00
  

Fix information

• Fixed component name

  IMS V7

• Fixed component ID

  5655B0100

Applicable component levels

• R700 PSY UK31813

     UP07/12/06 P F712