PH39536: TCPIP NOT PROCESSING XCF MESSAGES FROM OTHER SYSPLEX MEMBERS

APAR status

• Closed as program error.

Error description

• TCPIP encounters an ECSA shortage either at the system level or
  due to an ECSALIMIT configured in the TCPIP profile.  If the
  function that processes incoming XCF messages from other members
  of the sysplex is scheduled at this time it may never begin
  execution.  When this happens a gate flag is left on that
  prevents a new instance of the message processor from being
  started.  New incoming XCF messages will remain on the queue and
  never be processed.
  The XCF messages provide the dynamic VIPA state of other TCPIP
  stacks as well as information about sysplex distributed
  connections.  When these messages are not processed it is
  possible for a dynamic VIPA to exist this TCPIP stack and
  another member of the sysplex.  This can cause a disruption to
  the dynamic VIPA traffic.
  
  Messages related to the ECSA storage shortage and the TCPIP
  stack leaving the sysplex.
  EZZ4360I tcpip ECSA CONSTRAINED
  EZZ4361I tcpip ECSA CRITICAL
  EZZ9676E SYSPLEX PROBLEM DETECTION CLEANUP HAS SUCCEEDED FOR
  tcpip
  
  Messages issued when TCPIP rejoins the sysplex successfully.
  This does not provide any indication whether the XCF messages
  can be processed or not.
  EZD1176I tcpip HAS SUCCESSFULLY JOINED THE TCP/IP SYSPLEX GROUP
  EZBTCPCS
  EZD1192I THE VIPADYNAMIC CONFIGURATION WAS SUCCESSFULLY RESTORED
  FOR tcpip
  EZD1214I INITIAL DYNAMIC VIPA PROCESSING HAS COMPLETED FOR tcpip
  
  If the display of the sysplex status on this TCPIP stack does
  not show the dynamic VIPAs owned by other TCPIP stacks in the
  sysplex then XCF messages are not being processed.
  D TCPIP,tcpip,SYSPLEX,VIPADYN
  

Local fix

• RECOVERY ACTION:
  Recycle of the TCPIP stack is required to clear the gate flag
  and allow XCF messages to be processed.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Users of Sysplex and XCF messaging facility on release V2R5  *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * After a stack rejoins the sysplex (following a sys           *
  * autonomics problem detection, ie, ECSA storage critical), it *
  * fails to process incoming XCF messages from the other        *
  * members of the sysplex. This prevents the newly joined stack *
  * from processing sysplex workload (sysplex distribution for   *
  * example). This occurs when the flag indicating scheduling of *
  * XCF messages is left on from when the stack left the sysplex *
  * and prevents new scheduling after rejoin.                    *
  *                                                              *
  * The XCF messages are added to a message queue. Normally the  *
  * message processing routine would check for new messages      *
  * being added to the queue. If a TCP stack leaves the group at *
  * certain points during this processing, a flag can be         *
  * unintentionally left on that prevents new messages from      *
  * being scheduled.                                             *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply the PTF                                                *
  ****************************************************************
  TCPIP encounters an ECSA shortage either at the system level or
  due to an ECSALIMIT configured in the TCPIP profile. If the
  function that processes incoming XCF messages from other members
  of the sysplex is scheduled at this time it may never begin
  execution. When this happens a gate flag is left on that
  prevents a new instance of the message processor from being
  started.  New incoming XCF messages will remain on the queue and
  never be processed. The XCF messages provide the dynamic VIPA
  state of other TCPIP stacks as well as information about sysplex
  distributed connections. When these messages are not processed
  it is possible for a dynamic VIPA to exist this TCPIP stack and
  another member of the sysplex. This can cause a disruption to
  the dynamic VIPA traffic.
  
  Messages related to the ECSA storage shortage and the TCPIP
  stack leaving the sysplex.
  EZZ4360I tcpip ECSA CONSTRAINED
  EZZ4361I tcpip ECSA CRITICAL
  EZZ9676E SYSPLEX PROBLEM DETECTION CLEANUP HAS SUCCEEDED FOR
  tcpip
  
  Messages issued when TCPIP rejoins the sysplex successfully.
  This does not provide any indication whether the XCF messages
  can be processed or not.
  EZD1176I tcpip HAS SUCCESSFULLY JOINED THE TCP/IP SYSPLEX GROUP
  EZBTCPCS
  EZD1192I THE VIPADYNAMIC CONFIGURATION WAS SUCCESSFULLY RESTORED
  FOR tcpip
  EZD1214I INITIAL DYNAMIC VIPA PROCESSING HAS COMPLETED FOR tcpip
  
  If the display of the sysplex status on this TCPIP stack does
  not show the dynamic VIPAs owned by other TCPIP stacks in the
  sysplex then XCF messages are not being processed.
  D TCPIP,tcpip,SYSPLEX,VIPADYN
  

Problem conclusion

• Code has been updated to now reset the message scheduling flag
  and free any messages on the message queue when the stack leaves
  the sysplex group both due to error or normal processing.
  

Temporary fix

Comments

APAR Information

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

  PH39244

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

Fix information

• Fixed component name

  TCP/IP MVS

• Fixed component ID

  5655HAL00

Applicable component levels