Defining system objects for IBM MQ for z/OS

IBM MQ for z/OS® requires additional predefined objects for publish/subscribe applications, cluster, and channel control and other system administration functions.

The system objects required by IBM MQ for z/OS can be divided into the following categories:


Publish/subscribe objects

There are several system objects that you need to define before we can use publish/subscribe applications with IBM MQ for z/OS. Sample definitions are supplied with IBM MQ to help you define these objects. These samples are described in CSQ4INSG.

To use publish/subscribe you need to define the following objects:

  • A local queue called SYSTEM.RETAINED.PUB.QUEUE, which is used to hold a copy of each retained publication in the queue manager. Each full topic name could have up to one retained publication stored on this queue. If our applications will make use of retained publications on many different topics, or if your retained publication messages are large messages, the requirements for storage for this queue should be carefully planned, including assigning it to its own page set if the storage requirements for it are large. To improve performance, you should define this queue with an index type of MSGID (as shown in the supplied sample queue definition).
  • A local queue called SYSTEM.DURABLE.SUBSCRIBER.QUEUE, which is used to hold a persistent copy of the durable subscriptions in the queue manager. To improve performance, you should define this queue with an index type of CORRELID (as shown in the supplied sample queue definition).
  • A local queue called SYSTEM.DURABLE.MODEL.QUEUE, which is used as a model for managed durable subscriptions.
  • A local queue called SYSTEM.NDURABLE.MODEL.QUEUE, which is used as a model for managed non-durable subscriptions.
  • A namelist called SYSTEM.QPUBSUB.QUEUE.NAMELIST, which contains a list of queue names monitored by the queued publish/subscribe interface.
  • A namelist called SYSTEM.QPUBSUB.SUBPOINT.NAMELIST, which contains a list of topic objects used by the queued publish/subscribe interface to match topic objects to subscription points.
  • A topic called SYSTEM.BASE.TOPIC, which is used as a base topic for resolving attributes.
  • A topic called SYSTEM.BROKER.DEFAULT.STREAM, which is the default stream used by the queued publish/subscribe interface.
  • A topic called SYSTEM.BROKER.DEFAULT.SUBPOINT, which is the default RFH2 subscription point used by the queued publish/subscribe interface.
  • A topic called SYSTEM.BROKER.ADMIN.STREAM, which is the admin stream used by the queued publish/subscribe interface.
  • A subscription called SYSTEM.DEFAULT.SUB, which is a default subscription object used to provide default values on DEFINE SUB commands.


System default objects

System default objects are used to provide default attributes when you define an object and do not specify the name of another object to base the definition on.

The names of the default system object definitions begin with the characters "SYSTEM.DEFAULT" or "SYSTEM.DEF." For example, the system default local queue is named SYSTEM.DEFAULT.LOCAL.QUEUE.

These objects define the system defaults for the attributes of these IBM MQ objects:

  • Local queues
  • Model queues
  • Alias queues
  • Remote queues
  • Processes
  • Namelists
  • Channels
  • Storage classes
  • Authentication information

Shared queues are a special type of local queue, so when you define a shared queue, the definition is based on the SYSTEM.DEFAULT.LOCAL.QUEUE. You need to remember to supply a value for the Coupling Facility structure name because one is not specified in the default definition. Alternatively, you could define your own default shared queue definition to use as a basis for shared queues so that they all inherit the required attributes. Remember that you need to define a shared queue on one queue manager in the queue sharing group only.


System command objects

The names of the system command objects begin with the characters SYSTEM.COMMAND. You must define these objects before we can use the IBM MQ operations and control panels to issue commands to an IBM MQ subsystem.

There are two system command objects:
  1. The system-command input queue is a local queue on which commands are put before they are processed by the IBM MQ command processor. It must be called SYSTEM.COMMAND.INPUT, although an alias of SYSTEM.ADMIN.COMMAND.QUEUE for it can be defined, for compatibility with non-z/OS IBM MQ.
  2. SYSTEM.COMMAND.REPLY.MODEL is a model queue that defines the system-command reply-to queue.
There are two extra objects for use by the IBM MQ Explorer:

  • SYSTEM.MQEXPLORER.REPLY.MODEL queue
  • SYSTEM.ADMIN.SVRCONN channel

Commands are normally sent using nonpersistent messages so both the system command objects should have the DEFPSIST(NO) attribute so that applications using them (including the supplied applications like the utility program and the operations and control panels) get nonpersistent messages by default. If we have an application that uses persistent messages for commands, set the DEFTYPE(PERMDYN) attribute for the reply-to queue, because the reply messages to such commands are persistent.


System administration objects

The names of the system administration objects begin with the characters SYSTEM.ADMIN.

There are seven system administration objects:

  • The SYSTEM.ADMIN.CHANNEL.EVENT queue
  • The SYSTEM.ADMIN.COMMAND.EVENT queue
  • The SYSTEM.ADMIN.CONFIG.EVENT queue
  • The SYSTEM.ADMIN.PERFM.EVENT queue
  • The SYSTEM.ADMIN.QMGR.EVENT queue
  • The SYSTEM.ADMIN.TRACE.ROUTE.QUEUE queue
  • The SYSTEM.ADMIN.ACTIVITY.QUEUE queue


Channels queues

To use distributed queuing, you need to define the following objects:

  • A local queue with the name SYSTEM.CHANNEL.SYNCQ, which is used to maintain sequence numbers and logical units of work identifiers (LUWID) of channels. To improve channel performance, you should define this queue with an index type of MSGID (as shown in the supplied sample queue definition).
  • A local queue with the name SYSTEM.CHANNEL.INITQ, which is used for channel commands.

We cannot define these queues as shared queues.


Cluster queues

To use IBM MQ clusters, you need to define the following objects:

  • A local queue called the SYSTEM.CLUSTER.COMMAND.QUEUE, which is used to communicate repository changes between queue managers. Messages written to this queue contain updates to the repository data to be applied to the local copy of the repository, or requests for repository data.
  • A local queue called SYSTEM.CLUSTER.REPOSITORY.QUEUE, which is used to hold a persistent copy of the repository.
  • A local queue called SYSTEM.CLUSTER.TRANSMIT.QUEUE, which is the transmission queue for all destinations in the cluster. For performance reasons, you should define this queue with an index type of CORRELID (as shown in the sample queue definition).

These queues typically contain large numbers of messages.

We cannot define these queues as shared queues.


Queue sharing group queues

To use shared channels and intra-group queuing, you need to define the following objects:

  • A shared queue with the name SYSTEM.QSG.CHANNEL.SYNCQ, which is used to hold synchronization information for shared channels.
  • A shared queue with the name SYSTEM.QSG.TRANSMIT.QUEUE, which is used as the transmission queue for intra-group queuing. If you are running in a queue sharing group, you must define this queue, even if you are not using intra-group queuing.


Storage classes

You are recommended to define the following six storage classes. You must define four of them because they are required by IBM MQ. The other storage class definitions are recommended because they are used in the sample queue definitions.

    DEFAULT (required)
    This storage class is used for all message queues that are not performance critical and that don't fit in to any of the other storage classes. It is also the supplied default storage class if we do not specify one when defining a queue.

    NODEFINE (required)
    This storage class is used if the storage class specified when you define a queue is not defined.

    REMOTE (required)
    This storage class is used primarily for transmission queues, that is, system related queues with short-lived performance-critical messages.

    SYSLNGLV
    This storage class is used for long-lived, performance-critical messages.

    SYSTEM (required)
    This storage class is used for performance critical, system related message queues, for example the SYSTEM.CHANNEL.SYNQ and the SYSTEM.CLUSTER.* queues.

    SYSVOLAT
    This storage class is used for short-lived, performance-critical messages.

We can modify their attributes and add other storage class definitions as required.


Defining the system object dead-letter queue

The dead-letter queue is used if the message destination is not valid. IBM MQ puts such messages on a local queue called the dead-letter queue. Although having a dead-letter queue is not mandatory, you should regard it as essential, especially if you are using distributed queuing or one of the IBM MQ bridges.

Do not define the dead-letter queue as a shared queue. A put to a local queue on one queue manager might get put to the dead letter queue. If the dead letter queue was a shared queue, a dead letter queue handler on a different system could process the message and put it on a queue with the same name, but because this is on a different queue manager, it would be the wrong queue, or have a different security profile. If the queue did not exist, it would fail to reprocess it.

If you decide to define a dead-letter queue, you must also tell the queue manager its name. To do this use the ALTER QMGR DEADQ(queue-name) command. For more information see Altering queue manager attributes.


Default transmission queue

The default transmission queue is used when no other suitable transmission queue is available for sending messages to another queue manager. If you define a default transmission queue, you must also define a channel to serve the queue. If we do not do this, messages that are put on to the default transmission queue are not transmitted to the remote queue manager and remain on the queue.

If you decide to define a default transmission queue, you must also tell the queue manager its name. To do this use the ALTER QMGR command.


Internal queues

  • Pending data queue

    • A queue defined for internal use, SYSTEM.PENDING.DATA.QUEUE, supports the use of durable subscriptions in a JMS publish/subscribe environment.
  • JMS 2.0 delivery delay staging queue

    • If the delivery delay functionality provided by JMS 2.0 is used then an internal staging queue, SYSTEM.DDELAY.LOCAL.QUEUE, must be defined. This queue is used by the queue manager to temporarily store messages sent with a non-zero delivery delay until the delivery delay is completed, and the message is put to its target destination. A sample definition for this queue is provided, commented out, in CSQ4INSG.
    • When you define the SYSTEM.DDELAY.LOCAL.QUEUE queue, you must set the STGCLASS, MAXMSGL and MAXDEPTH attributes for the anticipated number of messages that will be sent with a delivery delay. Additionally when defining the SYSTEM.DDELAY.LOCAL.QUEUE queue ensure that only the queue manager can put messages to this queue. Care should be taken to ensure that no user identifier has the authority to put messages to this queue.


Channel authentication queue

For internal use of channel authentication the SYSTEM.CHLAUTH.DATA.QUEUE queue is required. Sample definitions are supplied with IBM MQ to help you define these objects. This sample is described in CSQ4INSA, which also defines some default rules.