Recovery during restart

IMS can recover from many types of failures during restart. In some cases, the recovery is automatic, for example, backout of transactions and the associated database updates. In other cases, you request a type of recovery on the restart command, for example, you can tell IMS to rebuild the message queues from a previous backup copy.

Generally, the way you restart IMS corresponds to the way you shut it down, or the way it fails.

If you only shut down part of IMS (for example, take a database offline or detach a line), you need to restart only that part.
If you shut down all of IMS or if IMS fails, you need to restart the whole IMS subsystem.

After recovering a failed component, you can restart it. For example, if you take a database offline because of an I/O error and then recover it, you can again make it available to applications.

To restart only part of IMS, use a command that starts the component that is stopped.

Before you can restart the entire IMS subsystem, you must first perform all necessary recovery. You can then restart IMS in one of three ways:

Normal restart initializes an IMS subsystem that has not failed. A normal restart can be either a cold start which restarts IMS without reference to any previous execution of IMS, or a warm start which restarts an IMS subsystem that was terminated using a /CHECKPOINT command. The normal restart command is /NRESTART.
Emergency restart initializes an IMS subsystem that has failed. During an emergency restart, IMS resets transactions and active regions, and restores databases and message queues to the most recent sync point. You must manually restart batch and batch message processing (BMP) regions. The emergency restart command is /ERESTART.
Automatic restart reduces MTO intervention and can make restart faster because IMS automatically chooses the appropriate restart command. The operator does not enter a restart command; instead you specify automatic restart by coding AUTO=Y in the JCL for the IMS control region.