/ON=(node[,...])

    In an OpenVMS Cluster, specifies the nodes on which a clusterwide
    queue manager can run. The default value for the node list is
    the asterisk (*)  wildcard character, meaning that all nodes in
    the cluster are eligible to run the queue manager. If the node on
    which the queue manager is running leaves the cluster, the queue
    manager can automatically fail over to any available node in the
    cluster. However, to specify a preferred order in which the nodes
    should claim the queue manager, or to limit the nodes which can
    run it, you must specify the /ON qualifier.

    The node list you specify is stored in the queue database.
    Anytime the START/QUEUE/MANAGER command is entered and neither
    the /NEW_VERSION nor /ON qualifier is specified, the /ON list
    stored in the queue database remains unchanged.

    For highest availability, specify the asterisk (*)  wildcard
    character as the last node in the node list to indicate that
    any remaining unlisted node can claim the queue manager, with
    no preferred order. If you do not specify the asterisk (*)
    wildcard character last in the node list, the queue manager can
    only fail over if one of the nodes in the list is available;
    however, if you want to exclude certain nodes from being eligible
    to run the queue manager, you cannot use the asterisk (*)
    wildcard character. You cannot specify the asterisk (*)  wildcard
    character as part of a node name.

    Anytime the START/QUEUE/MANAGER command is entered (with or
    without the /ON qualifier), the job controller will check to
    see if one or more preferred queue manager nodes was currently
    or previously specified with the /ON qualifier. If one or more
    preferred nodes was specified, and the queue manager is running
    on a node other than the first available node of those specified,
    the queue manager process is moved from its current node and
    restarted on the first available preferred node. Despite the
    transition, queues on the running nodes are not stopped. All
    requests to the queuing system, for example, PRINT, SUBMIT, and
    SHOW ENTRY requests, will complete as expected.