IBM Support

OA60274: A SYSZVVDS DEADLOCK MAY OCCUR IF PERFORMING DFSMSHSM RECALLS FROM MULTIPLE SYSTEMS DIRECTED TO THE SAME TARGET VOLSER

A fix is available

Subscribe

You can track all active APARs for this component.

 

APAR status

  • Closed as program error.

Error description

  • When multiple HSM RECALLs occur simultaneously from multiple
    systems, and they are directed to the same output volume, a
    small timing window exists that may allow a SYSVVDS deadlock to
    occur.
    For example:
    RECALL #1 on System A will read the target VVR for update and
    obtain serialization. In the small window before RECALL #1
    attempts to update the target VVR, RECALL #2 makes the same read
    request from System B.  When RECALL #1 attempts to update the
    VVR, it cannot obtain the proper serialization because of
    actions being performed by RECALL #2.  RECALL #2 cannot proceed
    because of the serialization held by RECALL #1, thus causing a
    deadlock.
    The timing window with the events occurring in such a manner is
    very small.
    
    KNOWN IMPACT:
    This may cause ENQUE HANG for SYSZVVDS, thus causing some
    processes not to complete.
    
    ADDITIONAL SYMPTOMS:
    HANG
    WAIT
    VVDs
    VVR
    ENQUEUE
    
    msgarc1170I rsn07 during recall of GDG generations and
    MSGADR497e rc50 rsn06 when attempting to update the VVR/NVR's
    class transition date.  This indicates the VVR/NVR does not
    exist even though DFSMSdss created it successfully.  Data set
    is recataloged on the volume as well msgarc0950I error invoking
    catalog delete rc50 rsn06 and msgarc0937I failed to scratch SMS
    managed data set.
    Application then attempts to open the data set which fails
    with msgiec143I 213-2C abend213 rc2c.
    
    CATKEYS: CAT2020 CATHANG CATDSS CATHSM CATVVDS
    

Local fix

  • Catalog's Contention Detection Feature can be used to identify
    SYSZVVDS deadlocks, and automatically redrive a hung catalog
    task. This can be implemented with the Catalog Modify command:
    
    F CATALOG,CONTENTION(resource,wait_time,action-flags)
    
    Where resource = SYSZVVDS; wait_time = time in minutes to wait
    before taking action; and action-flags = NR which indicates
    'notify & redrive'.
    
    This can also be permanently implemented in the IGGCATxx parmlib
    member with:
    
    resource(minutes,action)
    
    where resource = SYSZVVDS; minutes = time in minutes to wait
    before taking action; and action = NR
    

Problem summary

  • ****************************************************************
    * USERS AFFECTED:                                              *
    * All users of DFSMShsm RECALL and DFSMSdss RESTORE that       *
    * simultaneously RECALL or RESTORE multiple data sets that are *
    * directed to the same output volume.                          *
    ****************************************************************
    * PROBLEM DESCRIPTION:                                         *
    * An application was triggering multiple HSM data set RECALLs  *
    * across multiple instances of HSM. SMS directed the data sets *
    * to the same volume. During these RECALLs a deadlock occurred *
    * on the SYSZVVDS resource of the target volume.               *
    ****************************************************************
    * RECOMMENDATION:                                              *
    ****************************************************************
    

Problem conclusion

  • DFSMSdss has been modified to correct a serialization issue that
    caused a small timing window allowing the ENQUEUE deadlock
    during DFSMShsm RECALL.
    

Temporary fix

Comments

APAR Information

  • APAR number

    OA60274

  • Reported component name

    DFSMSDSS, ISMF

  • Reported component ID

    5695DF175

  • Reported release

    240

  • Status

    CLOSED PER

  • PE

    NoPE

  • HIPER

    NoHIPER

  • Special Attention

    NoSpecatt / Xsystem

  • Submitted date

    2020-10-02

  • Closed date

    2021-03-04

  • Last modified date

    2021-05-07

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

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

    UJ05016 UJ05017

Modules/Macros

  • ADRRDVVR ADRSB212 ADRSB114 ADRMFAMS ADRDALOC ADRCNVSM ADRPREVS
    ADRTDLOG
    

Fix information

  • Fixed component name

    DFSMSDSS, ISMF

  • Fixed component ID

    5695DF175

Applicable component levels

  • R240 PSY UJ05017

       UP21/03/16 P F103

  • R230 PSY UJ05016

       UP21/03/16 P F103

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.

[{"Business Unit":{"code":"BU054","label":"Systems w\/TPS"},"Product":{"code":"SG19M","label":"APARs - z\/OS environment"},"Platform":[{"code":"PF054","label":"z\/OS"}],"Version":"240"}]

Document Information

Modified date:
08 May 2021