PM52121: IMS TYPE 47 RECORD MAY OVERLAY CSA STORAGE IF BMP UPDATES MORE THAN 300 FULL FUNCTION DATABASES + HALDB PARTITIONS.

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• If a BMP Region updates more than about 300 Full Function
  Databases + HALDB Partitions and the BMP Region does not
  take a checkpoint prior to the IMS Control Region taking
  a checkpoint, an overlay of CSA storage may occur due to
  the Type47 Log Record overlaying storage beyond the end
  of the Log Work Area.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: This apar affects all IMS V11 customers      *
  *                 running BMP regions and BATCH IMS regions    *
  *                 that make updates to large numbers of        *
  *                 databases (242+) during a single commit      *
  *                 cycle.                                       *
  ****************************************************************
  * PROBLEM DESCRIPTION: ABEND0C4, ABEND0Cx, overlays, and       *
  *                      unpredictable results may occur when    *
  *                      attempting to write IMS type47 log      *
  *                      records that include BMP regions or IMS *
  *                      BATCH regions that have made updates to *
  *                      more than 242 individual databases      *
  *                      without committing the data.            *
  ****************************************************************
  * RECOMMENDATION: INSTALL CORRECTIVE SERVICE FOR APAR/PTF      *
  ****************************************************************
  Type47 log records are written at the end of IMS checkpoints and
  include region data and a list of changed databases that have
  not yet been committed for each IMS BMP region.  When the list
  of databases exceeds the capacity of the type47 log area in
  which they are built, overlays occur when DFSRCP10 attempts to
  build them.  Each partition of a HALDB is considered an
  individual database and has magnified the issue.
  
  The overlays can result in an ABEND0C4 referencing protected
  storage, but may also overlay storage that is not protected and
  cause unpredictable problems when the real storage owner
  attempts to use corrupted data.
  
  An additional problem was detected in which type0408 records may
  overlay storage using the same changed database list that was
  used in the type47 record.  The type0408 records have larger
  data base entries than the entries for the type47 record.  This
  limits the size of changed database list to what will fit into a
  type0408 record built in the PSTLOGWA, before an overlay occurs.
  That limit is reached at 242 databases.
  

Problem conclusion

• POSTREQ PM73179 PM79248
  GEN:
  KEYWORDS:
  
  *** END IMS KEYWORDS ***
  To prevent the dependent regions from altering more databases
  than can be built into a type0408 or type47 log record,
  BMP regions and BATCH IMS regions will be terminated ABENDU0199
  if they attempt database updates after reaching the maximum
  changed database count.  This maximum can be estimated by using
  the following decimal formula:
  
  Let
   N= the decimal number of 1024 byte blocks allocated for
      IMS Log Data Set
  
  Then
   Changed Database Limit = (64N - 10) Databases
  
  As a fail safe to prevent the overlays while building the log
  records, ABENDU0109 will be issued to terminate IMS if a record
  will exceed the capacity of the largest IMS Log Record that can
  be built, based on the IMS Log Dataset allocation (i.e. IMS log
  block size).
  
  However, abending the dependent regions with ABENDU0199 prior
  to getting changed database lists that are too large to log
  should prevent the ABENDU0109 from occurring.
  
  CODE CHANGES:
  
  IPST
   Added new field PSTDBCNT to contain the DB change list count.
  
  ISCD
   Added new field SCDDBCNT to hold the maximum count of database
   names that will fit into a type0408 record using the size of
   the IMS LOG Dataset, defined by the IMS account.
  
  DFSFDLS0
   Initialize the new SCD field SCDDBCNT.  This is accomplished by
   using SCDLLOGL (max log record size based on the IMS log
   allocation size, i.e. IMS Log block size), stripping off the
   type0408 record overhead size, and determining how many
   database names will fit in the remaining area.
  
  DFSXLGI0
   Initialize the new SCD field SCDDBCNT in a BATCH IMS System.
   This is accomplished by using SCDLLOGL (max log record size
   based on the IMS Log allocation size, i.e. IMS Log block size),
   stripping off the type0408 record overhead size, and
   determining how many database names will fit in the remaining
   area.
  
  DFSRDBL0
   Update the new PST field, PSTDBCNT, for each new database added
   to the changed database list.
  
  DFSDLA00
   ABENDU0199 when attempting a DB altering call when the limit
   of the changed database list has been reached.
  
  DFSFXC50
   Reset PSTDBCNT when Phase2 syncpoint has been called.
  
  DFSRCP10
   Monitor storage while building type47 records.  Acquire new
   storage when the changed database count exceeds the number that
   can be logged in the present buffer.
  
   ABENDU0109-01 when the record cannot be built in the largest
   IMS Log record based on the IMS Log Dataset allocation size
   (i.e. IMS Log block size).
  
   ABENDU0109-02 when storage cannot be acquired to build type47
   records.
  
  DFSRBOI0
   Added code to acquire and release storage to build a list of
   databases that fail dynamic backout and must be presented to
   DBRC. The dynamic storage is also used in building the
   type0408 log records.
  
  DFSST500
   Process type0408 log records on a Tracking IMS System when
   the database list within it will cause overlays building
   the DBRC VERUOR or UPDUOR list in the PSTLOGWA. This is
   accomplished acquiring and releasing dynamic storage when
   needed.
  
  DFSRBLB0
   Corrected ERE NOBMP and ERE COLDBASE paths to handle large
   Type0408 log records and SAVUOR calls. Handled these by
   processing single UORs at a time. Since the changed DB
   count is limited to the amount that will fit into a Type0408
   maximum size log record, there is no problem writing these
   using the largest logsize, i.e. IMS Log block size (SCDLLOGL).
  
  The IMS MESSAGE AND CODES MANUAL will have the following abends
  documented:
  
  U0199
  Explanation
    A BMP or IMS BATCH region has exceeded the count of databases
    that can be altered within a commit cycle.
  
  Programmer Response
    Insure that BMP or IMS BATCH regions do not make alterations
    to more databases than will fit in the largest log record
    allocated for the IMS Log Data Set (i.e. IMS Log block size)
    without issuing a syncpoint request or increase the IMS Log
    Data Set size to accommodate the number of changed databases.
  
    An estimate of the maximum number of changed databases that
    will fit into the largest log record can be determined using
    the following decimal formula:
  
      Let
       N= the decimal number of 1024 byte blocks allocated for
          IMS Log Data Set (i.e. IMS Log block size)
      Then
       Changed Database Limit = (64N - 10) Databases
  
  System Action
    The BMP or Batch Region is abended U0199.
  
  Operator Response
    After correcting the problem, restart the BMP or IMS BATCH
    region.
  
  Module
    DFSECP20
  
  
  U0109
  Explanation
    The IMS Control Region has been abended while attempting to
    build a type47 log record.
  
   Reason Code Explanation
    R15=1
     A type47 log record cannot be created because the changed
     database list of a BMP exceeds the capacity of the log record
     in which it is to be built.
  
    R15=2
     An IMODULE GETMAIN request failed to acquire storage to build
     a type47 log record.
  
  Analysis
   R15=1
    This abend should not occur but is added to prevent overlays
    in the event that BMP or IMS BATCH regions have corrupted
    database lists that will not fit into IMS log records.
  
   R15=2
    An IMODULE GETMAIN failed while attempting to acquire storage
    to build the type47 log record.  R10 contains the IMODULE
    GETMAIN return code.
  
  System Action
   IMS is abended with ABENDU0109
  
  Operator Response
   After correcting the problem, restart IMS with /ERE.
  
  System Programmer Action
   Contact IBM@Software Support
  
  Module
   DFSRCP10
  

Temporary fix

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

Comments

• PINNED P12/09/18 (ATXT) TO ADD POSTREQ PM73179 INFO.
   **** PE12/09/18 PTF IN ERROR. SEE APAR PM73179 FOR DESCRIPTION.
  PINNED P12/12/14 (ATXT) TO ADD POSTREQ PM79248 INFO.
   **** PE12/12/14 PTF IN ERROR. SEE APAR PM79248 FOR DESCRIPTION.
  PINNED P14/02/28 (ATXT) TO ADD POSTREQ PI11660 INFO.
  **** PE14/02/28 PTF IN ERROR. SEE APAR PI11660 FOR DESCRIPTION.
  ×**** PE14/02/12 FIX IN ERROR. SEE APAR PI11660  FOR DESCRIPTION
  

APAR Information

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

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

  PM63331 UK78835

Modules/Macros

• DFSDLA00 DFSFDLS0 DFSFXC50 DFSRBLB0 DFSRBOI0
  DFSRCP10 DFSRDBL0 DFSST500 DFSXLGI0 IPST     ISCD
  

Fix information

• Fixed component name

  IMS V11

• Fixed component ID

  5635A0200

Applicable component levels

• R100 PSY UK78835

     UP12/05/19 P F205 Ž

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.