z/OS structure duplexing for CQS

Structure duplexing is an optional z/OS-managed process for failure recovery of queue structures and resource structures. In this process, z/OS® creates a duplex copy of a structure in advance of a failure, then maintains the structures in a duplexed state during normal operation.

If a queue structure fails and duplexing is enabled, z/OS switches to the unaffected structure instance. If a queue structure fails and duplexing is not enabled, CQS rebuilds the structure based on data from the most recent checkpoint and z/OS log entries. The advantage of duplexing queue structures in the event of a failure is in avoiding the overhead of a CQS-managed structure rebuild.

If duplexing is enabled and a resource structure fails, z/OS switches to the unaffected structure instance. If duplexing is not enabled and a resource structure fails, the data objects are lost because resource structures do not support checkpoint or logging. CQS repopulates the resource structure with control information. CQS notifies its clients to repopulate the structure. It is up to the clients to repopulate the resource structure if necessary.

Recommendation: Enable structure duplexing for resource structures.

If both instances of a structure fail at the same time, structure duplexing does not work and all data objects are lost. If the failed structure is a resource structure, the CQS client must repopulate it. If the failed structure is a queue structure, CQS recovers the structure using structure rebuild.

Structure duplexing is optional. To use it, you must enable the z/OS duplexing function. Perform the following steps to enable this function:

Procedure

  1. Ensure that the sysplex is defined as duplexing capable.
  2. Add the following parameter to your CFRM couple data set format utility: ITEM NAME(SMDUPLEX) NUMBER(1)
  3. Migrate to an environment in which system-managed duplexing is enabled from a CFRM standpoint. A nondisruptive migration of CFRM couple data sets is required. Only z/OS systems at a level that supports system-managed duplexing are capable of using system-managed CFRM couple data sets that are duplexing-capable. Therefore, take the following steps:
    1. Incrementally migrate all systems in the sysplex that are using CFRM to the z/OS level that supports system-managed duplexing.
    2. Format system-managed duplexing-capable CFRM couple data sets and bring them into use as the primary and alternate CFRM couple data sets for the configuration.
Important: After you enable z/OS duplexing, you cannot return to downlevel CFRM couple data sets (ones that are not system-managed duplexing-capable) without disruption. Doing so requires a sysplex-wide IPL of all systems using the system-managed duplexing-capable data sets.

After an uplevel CFRM couple data set is in use in the sysplex, system-managed duplexing can be started and stopped in a nondisruptive manner. To turn this function on or off, even while the CFRM couple data set is in use, modify the CFRM policy DUPLEX parameter or use the SETXCF START/STOP,REBUILD,DUPLEX operator command.

To enable system-managed duplexing for a particular structure, the structure must be defined as duplexing-capable. Defining a structure as duplexing capable also defines it as system-managed rebuild-capable. Add the following parameter to your CFRM active policy: 
DUPLEX (ENABLED)
or 
DUPLEX(ALLOWED)

If DUPLEX(ENABLED) is defined in the CFRM active policy, the system programmer or z/OS internally can initiate the duplexing rebuild. z/OS triggers the start of duplexing rebuild based on a timer or upon detection of certain events (such as connect, disconnect, and policy change). When CQS initializes and connects to a structure defined with DUPLEX(ENABLED), z/OS starts a duplexing rebuild.

If DUPLEX(ALLOWED) is defined in the CFRM active policy, the duplexing rebuild must be initialized by the system programmer using the following command: 

SETXCF START,REBUILD,DUPLEX,STRNAME=strname
Important: If you define overflow structures with DUPLEX(ENABLED), IMS initialization allocates the overflow structure and duplexing begins. If IMS initialization determines that the overflow structure is not needed, it deletes it and duplexing terminates. If you want to avoid this unnecessary overhead, during CQS initialization define the overflow structure with DUPLEX(ALLOWED) and initiate duplexing with a SETXCF command when the structure goes into overflow mode.

Once duplexing is established, the structure remains in that state indefinitely. Duplexing can be stopped internally by z/OS if an error occurs (such as link failure, structure failure, and CFRM policy change). The system programmer can explicitly stop duplexing using the following command, where you specify KEEP=OLD to keep the old structure and KEEP=NEW to keep the new structure.

SETXCF STOP,REBUILD,DUPLEX,STRNAME=strname,KEEP=OLD/NEW

Planned reconfiguration (such as a CFRM policy change or taking a coupling facility offline for maintenance) is supported. Structure rebuild is not permitted for a structure that has established duplexing, so the duplexing must be stopped first. Perform the following steps:

  1. Stop duplexing.

    Stop duplexing and switch the structure to simplex mode by issuing the following command: 

    SETXCF STOP,REBUILD,DUPLEX,STRNAME=strname,KEEP=OLD/NEW
  2. Reconfigure.

    Make the change required for planned reconfiguration.

  3. Initiate duplexing rebuild.

    Initiate a new duplexing rebuild by issuing the following command: 

    SETXCF START,REBUILD,DUPLEX,STRNAME=strname