IBM Support

PI15999: REBUILD SCA CAUSED SYSTEM TO HANG FROM RL LATCH CONTENTION BETWEEN DDF DBAT THREAD IN COMMIT AND SCA REBUILD SERVICE TASK

A fix is available

Subscribe

You can track all active APARs for this component.

APAR status

  • Closed as program error.

Error description

  • DB2DDF DB2D2PC DB2XA DB2DSHR DB2TERM defect pi15999 dpi15999
    Customer issues the SETXCF START,REBUILD,STRNAME=xxx_SCA
    command and the system hangs.  At the time of the command
    issuance, a DDF online transaction is going through commit
    processing and holding RL latch. DB2 service task that
    supports the SCA rebuild also is suspended waiting for
    the RL latch.  The DDF dbat thread is also waiting for the
    same DB2 service task to complete, thus the hang.
    DB2 needs to be recycled to break the hang.
    ***************************************************************
    Additional Symptoms and Keywords:
     ABEND026 AB026 S0026 026
      026-08118001 08118001 RC08118001
     RLLATCH SCA Rebuild
    

Local fix

Problem summary

  • ****************************************************************
    * USERS AFFECTED: All Distributed Data Facility (DDF) users.   *
    *                 Specifically where DB2 is a member of a      *
    *                 data sharing group.                          *
    *                 DB2 11 for z/OS users only.                  *
    ****************************************************************
    * PROBLEM DESCRIPTION: Deadlock between SCA Rebuild and an     *
    *                      XA two phase commit sequence from a     *
    *                      remote system.                          *
    *                      The SCA Rebuild deadlock condition      *
    *                      results in periodic abend 026-08118001  *
    *                      conditions.                             *
    ****************************************************************
    * RECOMMENDATION:                                              *
    ****************************************************************
    In order to support remote client XA transaction managers when
    DB2 is a member of a data sharing group, potential indoubt XID
    information must be recorded in the SCA since the remote XA
    client may contact ANY member of the group in order to obtain
    indoubt XID information related to ALL members.
    As a result, during DRDA XA Phase 1 Commit processing,
    xa_prepare(), the DB2 z/OS server will add an "xid" entry to
    the SCA.  Once a (commit/abort) decision is received, DB2 will
    remove the "xid" entry from the SCA.
    In DB2 11 for z/OS, DB2 DDF is now involved in SCA Rebuild
    processing.  A deadlock condition may occur if an SCA Rebuild
    is initiated while DB2 (DDF) is able to service DRDA XA
    transactions.
    The deadlock occurs because SCA Rebuild processing must wait
    for a latch that is held by another thread that is processing
    the XA transaction decision.  However, the XA transaction
    decision requires that the associated entry be deleted from the
    SCA but this waits due to the SCA Rebuild that is underway.
    This deadlock condition results in the periodic detection of
    abend 026-08118001.  The DB2 subsystem must be terminated to
    break the deadlock condition.
    

Problem conclusion

  • DB2 DDF SCA entry delete processing has been changed to
    eliminate the serialization issue that can lead to the deadlock,
    or to conditionally delete the SCA entry since (if an SCA
    Rebuild is underway) there is no need to delete the SCA entry
    because the SCA Rebuild will reflect the completion of the XA
    transaction.
    

Temporary fix

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

Comments

APAR Information

  • APAR number

    PI15999

  • Reported component name

    DB2 OS/390 & Z/

  • Reported component ID

    5740XYR00

  • Reported release

    B10

  • Status

    CLOSED PER

  • PE

    NoPE

  • HIPER

    YesHIPER

  • Special Attention

    NoSpecatt

  • Submitted date

    2014-04-14

  • Closed date

    2014-05-15

  • Last modified date

    2014-06-03

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

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

    UI18017

Modules/Macros

  • DSNLCMT2 DSNLTBAB DSNLTRD  DSNLTRSI DSN7LDE1
    

Fix information

  • Fixed component name

    DB2 OS/390 & Z/

  • Fixed component ID

    5740XYR00

Applicable component levels

  • RB10 PSY UI18017

       UP14/05/30 P F405 ½

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.



Document information

More support for: DB2 for z/OS

Software version: 11.0

Reference #: PI15999

Modified date: 03 June 2014