OA56231: STRUCTURE REBUILD HANG FOLLOWING LOSSCONN EVENT UNDER MSGBASED PROTOCOL

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Following a CF failure, structure rebuild hangs were encountered
  showing outstanding connector responsces as evident by D XCF,STR
  output when using CFLCRMGMT. In order to be exposed to
  this issue, there would have to be a loss of CF connectivity
  (IXC528I) and also the loss of a system in a close timeframe.
  The issue also involves a timing element to it.
  
  ANALYSIS:
  Following loss of CF connectivity, a CF lossconn event will
  drive structure rebuild activity.
  
  A problem was found where the CFRM message-based manager system
  may not free EMB control blocks following a LOSSCONN event
  signal.
  
  KNOWN IMPACT:
  Structure rebuilds hang.
  
  Issuing SETXCF STOP,MSGBASED followed by
  SETXCF START,MSGBASED addresses the rebuild hangs.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Users in a Parallel Sysplex environment                      *
  * with the optional XCF function CFLCRMGMT                     *
  * enabled.                                                     *
  *                                                              *
  * SYSPLEXDS                                                    *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * Due to timing or lack of XCF signaling                       *
  * connectivity, coupling facility (CF)                         *
  * structure connector termination (for                         *
  * example, disconnect) before the system                       *
  * is removed from the sysplex may cause                        *
  * the XCF MSGBASED manager system to                           *
  * retain residual control blocks if the                        *
  * system is removed from the sysplex and                       *
  * the struture is no longer allocated.                         *
  * The residual control blocks may cause                        *
  * CF LOSSCONN RECOVERY MANAGEMENT to                           *
  * pause, which will cause any XCF                              *
  * REALLOCATE process to hang.                                  *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Fix should be installed via rolling IPL.                     *
  * Full functionality of the fix will not                       *
  * be available until it is installed on                        *
  * every system in the sysplex. The fix for                     *
  * this APAR need only be installed on the                      *
  * system assigned the role of XCF CFRM                         *
  * MESSAGE-BASED MANAGER. However, when                         *
  * that system is removed from the sysplex,                     *
  * the role is reassigned to an active                          *
  * system in the sysplex. An IPL is                             *
  * required to activate this fix on each                        *
  * system of the SYSPLEX, however, a                            *
  * rolling IPL is sufficient to accomplish                      *
  * the activation.                                              *
  *                                                              *
  * Before the fix is installed on every                         *
  * system in the sysplex, system commands                       *
  * SETXCF STOP,MSGBASED and SETXCF                              *
  * START,MSGBASED can be used on a system                       *
  * with the fix installed to make the fix                       *
  * effective. The role of the XCF CFRM                          *
  * MESSAGE-BASED manager system is assumed                      *
  * by the system that successfully executes                     *
  * the SETXCF START,MSGBASED operator                           *
  * command.                                                     *
  ****************************************************************
  The XCF MSGBASED manager system incorrectly relies on a CF
  structure connector termination event signal to clean up
  event control blocks for a structure. Due to timing or lack
  of XCF signaling connectivity, such an event signal may not
  be delivered to the manager system when another system is
  removed from the sysplex. If the structure is not allocated
  during system termination, the manager system may not clean
  up residual control blocks. The residual control blocks cause
  the manager system to process CF LOSSCONN RECOVERY MANAGEMENT
  as if a system is hung - causing CF LOSSCONN RECOVERY
  MANAGEMENT to pause. When CF LOSSCONN RECOVERY MANAGEMENT is
  paused, REALLOCATE processing will hang.
  

Problem conclusion

• Corrected XCF system termination processing to clean up
  residual XCF MSGBASED manager control blocks for CF structures
  that are not allocated.
  

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:

  UJ01071 UJ01072 UJ01073

Modules/Macros

• IXCL2RSR
  

Fix information

• Fixed component name

  XCF

• Fixed component ID

  5752SCXCF

Applicable component levels

• R7A0 PSY UJ01073

     UP19/10/30 P F910 ¢

• R7B0 PSY UJ01071

     UP19/10/30 P F910 ¢

• R7C0 PSY UJ01072

     UP19/10/30 P F910 ¢

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.