OA63312: INCORRECTLY HANDLING THE BREAKING DUPLEX FOR SM DUPLEX STR IN CAUSING DXR139E IXLRT RC0C RSN0C540C06. LO 22/08/29 PTF PECHANGE

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• SM duplex str were established between CF1 and CF2. When CF1 was
  lost, the in-flight request toward the CF1 was detected with
  IFCC due to the
  lack of respond from CF1. Path validation toward CF1 were
  performed and also failed with IFCC since CF1 was lost.
  In-flight request toward the CF2 was detected with IFCC and path
  validation toward CF2 were successful to confirm CF2 was still
  available. Under the recovery on the CF2 thread, IXLE1REC
  concluded the breaking duplex condition but determined to give
  up the copy on CF2 instead on CF1. As the 'surviving' copy of
  str on the CF2 was dropped and CF1 was no longer accessible,
  applications suffered further delay as the str must be rebuild.
  Connector to the str could issue :
  DXR139E with RC0C RSN0C540C06 and DXR122E with ABENDU2025
  
  ANALYSIS:
  Confirm the CF which was lost by finding the IXC518I in the
  syslog or output of VERBX MTRACE.
  IXC577I to confirm the surviving copy on the CF name mentioned
  in the IXC518I. IXC579I for the copy on surviving CF.
  
  KNOWN IMPACT:
  Application faliure as can not acess the str ( RC0C with RSN
  0xxx0C06 with IXL request) which lead to application (IRLM) shut
  down
  
  VERIFICATION STEPS:
  If the dump with collector of the str is available. Check the
  output from
  CTRACE COMP(SYSXES) SUB((.....)) to locate Entry with following
  id :
  09100006 with ARWEMFID xx, TmsCctraced wiht 10
  070B0001
  070B0005 with McftMfid xx, TmsCCByte 10
  030D0001
  09060005 MFID xx TMSCCByte 10  for the lost CF or
  09060007 MFID xx for the surviving CF
  070B0008 ARWEMFID xx as surviving CF LocalRequestRetcode with
  00000FFF
  
  PE Information:
  
  Users affected
  
    Installations exploiting parallel sysplex with duplexed lock
    structures, using either the synchronous or asynchronous
    duplexing protocol, and having the following OA60275 PTFs
    installed:
  
    HBB77C0 UJ08016   HBB77B0 UJ08015   HBB77A0 UJ08025
    Note: HBB77D0 PTF UJ08017 is not PE.
  
  User Impact:
    OA60275 provides toleration support for coupling facilities
    (CFs) defined on z16 processors, which implement CFLEVEL 25.
    OA60275 provided the required toleration support but regressed
    subchannel recovery for interface control checks (IFCCs) when
    they occur during processing of specific lock structure
    commands at CFLEVELs below 25.  The regression causes z/OS to
    break duplexing for the structure, and may cause loss of the
    structure.
  
    Loss of structure occurs only when the break-duplexing causes
    failover to a CF that is inaccessible or that contains an
    instance of the structure that is no longer viable.  This is
    very uncommon except in test scenarios involving CF
    deactivation.
  
    Break-duplexing without loss of structure is inconvenient but
    does not impact normal workload execution.  In the absence of
    hardware issues, IFCCs will be rare in a normal configuration.
  

Local fix

• Install PTFs for OA63312.  Until then:
  
  o Avoid running tests involving the deactivation of a CF
    containing lock structures.
  
  o If break-duplexings are disruptive, consider running with
    lock structures in simplex mode.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Installations exploiting parallel sysplex at                 *
  * z/OS V2R2 (HBB77A0) and above, using                         *
  * synchronously- or asynchronously-duplexed                    *
  * lock structures, with any coupling facility                  *
  * (CF) below CFLEVEL 25 in the configuration.                  *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * Inappropriate break-duplexing and                            *
  * potential loss of structure when                             *
  * a command to a CF lock structure                             *
  * incurs an interface control check                            *
  * (IFCC).                                                      *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * An IPL is required to activate this fix on                   *
  * each system of the sysplex.  A rolling IPL                   *
  * is sufficient to activate the fix.                           *
  ****************************************************************
  For a specific subset of locking commands, an operation directed
  to a duplexed CF lock structure instance that resides in a CF
  below CFLEVEL 25 and incurs an IFCC will inappropriately break
  duplexing.  If this occurs, subchannel recovery processing
  recommends to CFRM that it retain the peer structure instance.
  If the peer instance is no longer viable, for example if the CF
  housing it has been deactivated, the structure is lost.
  
  SYSPLEXDS
  R3931/K
  D/T3931
  

Problem conclusion

• Subchannel recovery will not break duplexing for a duplexed
  locking command that IFCCs.  Instead, when appropriate, the
  command will be internally restarted.
  KEYWORDS: D/T3931 R3931/K
  

Temporary fix

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

Comments

APAR Information

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

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

  UJ08992 UJ08993 UJ08994 UJ08995

Modules/Macros

• IXLE1REC
  

Fix information

• Fixed component name

  CROSS SYS.EXT.S

• Fixed component ID

  5752SCIXL

Applicable component levels

• R7A0 PSY UJ08995

     UP22/08/17 P F208 «

• R7D0 PSY UJ08994

     UP22/08/17 P F208 «

• R7B0 PSY UJ08992

     UP22/08/17 P F208 «

• R7C0 PSY UJ08993

     UP22/08/17 P F208 «

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.