PM89391: THREAD DOING AUTOBIND MAY HANG IN DSNTABA AFTER BEING CANCELLED

A fix is available

Subscribe

You can track all active APARs for this component.

APAR status

  • Closed as program error.

Error description

  • Thread doing Autobind may hang in DSNTABA after being cancelled
    .
    SKPT
    

Local fix

Problem summary

  • ****************************************************************
    * USERS AFFECTED: All users of DB2 for zOS                     *
    *                                                              *
    ****************************************************************
    * PROBLEM DESCRIPTION: An agent may become hung in CSECT       *
    *                      DSNTABA if a -CANCEL THREAD command is  *
    *                      issued for that agent while it was      *
    *                      holding an IRLM lock.                   *
    ****************************************************************
    * RECOMMENDATION:                                              *
    ****************************************************************
    When a -CANCEL THREAD command is issued for an agent that is
    holding a SKPT p-lock, EDM abort processing will save that
    SKPT p-lock on a global EDM chain to be cleaned up during IRLM
    synchronization that occurs during general agent abort/
    deallocation processing.  Any new agent that requires the same
    SKPT p-lock may also try to purge SKPT p-locks on the global
    EDM chain that are leftover from previous threads.
    
    This leads to a race condition between IRLM synchronization
    process for a cancelled thread an a new DB2 agent that may be
    interested in the SKPT p-lock.  If the latter wins the race,
    IRLM synchronization processing in CSECT DSNTABA will
    incorrectly suspend waiting for a system task that its scheduled
    to free the lock.  This will never occur since this system task
    will not find this lock, resulting in a indefinite hang in the
    CSECT DSNTABA.
    
    This race condition only exists when SKPT p-lock is involved.
    

Problem conclusion

  • Remove the race condition by having EDM cancel processing use
    an agent specific chain to save those SKPT p-locks that need to
    be freed during the abort/deallocation process.
    

Temporary fix

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

Comments

APAR Information

  • APAR number

    PM89391

  • Reported component name

    DB2 OS/390 & Z/

  • Reported component ID

    5740XYR00

  • Reported release

    A10

  • Status

    CLOSED PER

  • PE

    NoPE

  • HIPER

    YesHIPER

  • Special Attention

    NoSpecatt

  • Submitted date

    2013-05-20

  • Closed date

    2013-09-24

  • Last modified date

    2013-11-04

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

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

    UK97873 UK97874 UK97875

Modules/Macros

  • DSNGEELC DSNGEFLC DSNTABA  DSNTADL
    

Fix information

  • Fixed component name

    DB2 OS/390 & Z/

  • Fixed component ID

    5740XYR00

Applicable component levels

  • RA10 PSY UK97873

       UP13/10/10 P F310 ½

  • RB10 PSY UK97874

       UP13/10/10 P F310 ½

  • R910 PSY UK97875

       UP13/10/11 P F310 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.



Rate this page:

(0 users)Average rating

Add comments

Document information


More support for:

DB2 for z/OS

Software version:

A10

Reference #:

PM89391

Modified date:

2013-11-04

Translate my page

Machine Translation

Content navigation