Restarting a server

All three types of CICS data-sharing server (temporary storage, coupling facility data tables, and named counters) support automatic restart using the services of the Automatic Restart Manager (ARM).

The servers also have the ability to wait during start-up, using an Event Notification Facility (ENF) exit, for the coupling facility structure to become available if the initial connection attempt fails.

The server normally registers at start-up with ARM, so that it will be restarted automatically if it fails, subject to any rules in the installation ARM policy. If ARM registration fails for any reason except for ARM being unavailable, the server cannot be started. If ARM is unavailable, the server starts normally but has to be restarted manually if it fails.

The servers recognize the ARM return code that indicates that the ARM couple data set has not been formatted for the current MVS system, as being equivalent to ARM being unavailable.

A server does not start if registration fails with return code 8 or above.

When a server starts up, if it is unable to connect to its structure because of some environmental error such as a structure failure, it automatically waits for the structure to become available, using the Event Notification Facility (ENF) to watch for events relating to its structure. This wait occurs before the cross-memory interface is enabled, so the server is not visible to client regions at this time and will appear to be unavailable. While it is waiting, the server can be cancelled using the MVS CANCEL command if it is no longer required.

If the server is running normally, but the coupling facility interface reports a loss of connectivity or a structure failure, the server immediately terminates itself. This disconnects it from the coupling facility, and terminates the server side of any current cross-memory connections from client regions. The server will normally be restarted immediately by the ARM, but will continue to be unavailable to client regions until the coupling facility structure is available again (possibly as a new empty instance of the structure).

An abrupt coupling facility failure such as a power failure may result in a loss of connectivity indication even though the structure has failed, because the operating system cannot determine the state of the structure in that case. This could prevent a new structure from being allocated until the operating system can determine the status of the existing structure, for example after the failed coupling facility has been successfully restarted. If it is certain that the old structure has been lost, but the system has not yet recognized the fact, the operator may be able to save some time by issuing the SETXCF FORCE command to delete the old structure, allowing the system to go ahead and create a new instance of the same structure in a different coupling facility.

You can find more information about automatic restart of coupling facility servers in ../../com.ibm.cics.ts.recrestart.doc/topics/dfht2iv.html#dfht2iv