PH25530: ABEND04E RC00E70005 DSNXEDS1 M199 OCCURS ON PREPARE OF DYNAMIC SQL STATEMENT IN A VERY ACTIVE SYSTEM

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• ABEND04E RC00E70005 DSNXEDS1 M199 occurs on PREPARE of dynamic
  SQL statement in a very active system.
  Problem criteria:
  Multiple threads race to PREPARE the same dynamic SQL statement
  concurrently. The winner of the race is inserted into the cache,
  but then is immediately removed, by EDM's LRU algorithm.
  The loser of the race (first loser if there are multiple), will
  then get this abend.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All Db2 12 for z/OS users of dynamic                         *
  * statement caching.                                           *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * ABEND04E RC00E70005 DSNXEDS1 M199                            *
  * occurs on PREPARE of dynamic                                 *
  * SQL in a very active system.                                 *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply corrective PTF when available                          *
  ****************************************************************
  ABEND04E RC00E70005 DSNXEDS1 M199 occurs on PREPARE of dynamic
  SQL when multiple threads race to do a full PREPARE of the
  same dynamic SQL statement concurrently.  The winner of the
  race is inserted into the cache.  The loser registers interest
  in the new cached statement.
  There is a window between the time the winner's statement is
  inserted into the cache and the time when the loser registers
  interest in the the cached statement.  If the winner's cached
  statement is invalidated and removed from the cache before the
  loser can register interest, the lost must re-drive full
  PREPARE.
  The root cause of the problem is that this re-prepare path
  does not clean up properly before restarting.
  The problem can seem intermittent because for it to occur,
  multiple full prepares of the same statement must be
  executing concurrently with the invalidation of the cached
  entries for those same statements.
  The likely reason for cache invalidation in this case
  is the EDM cache pool is full which causes EDM to remove
  older cached statements using its LRU algorithm.
  

Problem conclusion

• Db2 is modified to re-initialize the OP and INFO control
  blocks prior to calling the parser in the re-prepare path.
  Additional keywords:  SQLDYNSTMTCACHE
  

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:

  UI70977

Modules/Macros

• DSNXEDS1 DSNXEDSC
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RC10 PSY UI70977

     UP20/08/18 P F008

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.