The following considerations will help you determine the policy information you need for each batch job or started task that will register as an element of automatic restart management. The policy parameters you would use to override the defaults are shown in parentheses. For more information on these parameters, see Automatic restart management parameters for administrative data utility.

1. Determine which batch jobs and started tasks will be using automatic restart management for recovery purposes. For the IBM® products that use automatic restart management, read their documentation for any policy considerations.
2. Determine if any of these elements is interdependent — that is, need to run on the same system. (RESTART_GROUP).

   Note that any elements that are not explicitly assigned to a restart group become part of the restart group named DEFAULT. Thus, if these elements are restarted, they are restarted on the same system.

3. Determine if there is an order in which MVS™ should restart these elements — if any elements in the restart group are dependent upon other elements being restarted and ready first (RESTART_ORDER).
4. Determine if the elements in a restart group need to be restarted at specific intervals (RESTART_PACING).
5. Determine if the element should be restarted when only the element fails, or when either the element or the system fails (TERMTYPE).
6. Determine whether specific JCL or command text is required to restart an element (RESTART_METHOD).
7. Determine if a minimum amount of CSA/ECSA is needed on the system where the elements in a restart group are to be restarted (FREE_CSA).
8. Determine if you want the elements in the restart group to be restarted on a specific system (TARGET_SYSTEM).
   Consider requirements like:

   • Which systems have access to the ARM couple data sets.
   • Which systems run within the same JES XCF group.
   • The workloads of the systems in the sysplex, and how they might be affected if these jobs were restarted.
   • CPU capacity needed.
   • DASD requirements.
   • Which systems have the class of initiator required by batch jobs that might need to be restarted on another system.
9. Determine if an element should be restarted and, if so, how many times it should be restarted within a given interval (RESTART_ATTEMPTS). If an element should not be restarted, set RESTART_ATTEMPTS to 0.
10. Determine how long MVS should wait for an element to re-register once it has been restarted (RESTART_TIMEOUT).
11. Determine how long MVS should wait for an element to indicate it is ready to work after it has been restarted (READY_TIMEOUT).