Resynchronizing sessions

To maintain the integrity of recoverable resources, messages, and queues in IMS across both subsystem and session failures, both half sessions must maintain the session information required for the resynchronization process.

Message integrity cannot be maintained without user intervention when either or both subsystems incur an error or when a user restart procedure destroys this information.

Resynchronization is required when it is possible for a recoverable work unit to be indoubt on a flow (primary-to-secondary or secondary-to-primary).

Definition: A work unit includes all transmissions between sync points within a bracket, as illustrated in the following figure. A sync point might have been requested by one or both half sessions without having been acknowledged. These conditions can be caused by a subsystem or session failure or an abnormal completion of a shutdown sequence.

Performing message resynchronization is unnecessary following a normal shutdown sequence (unless nonnegotiated BIND was sent), because both half sessions can come to a controlled, mutual understanding that no additional normal message traffic or sync-point requests are to occur prior to session termination. Message resynchronization is always required following nonnegotiable BIND to allow error conditions detected by the secondary half session to be communicated to the primary half session.

The half-session pairs resynchronize with the VTAM® BIND, set-and-test-sequence-numbers (STSN), and start data traffic (SDT) commands. The STSN command allows both half sessions to reestablish sync-point information (session sequence numbers), which is being maintained by both half sessions.

When message resynchronization is necessary, it must be completed successfully before either half session can resume normal data transmission.

The following figure shows two work unit examples.

Figure 1. Work unit examples
On Subsystem A, arrows point from RQ*1, RQ*1, and RQD2,CD in Work unit 1 and RQ*1,CD in Work unit 2 to Subsystem B. On Subsystem B, arrows point from RQ*1 and RQD2 (CD or EB optional) to Subsystem A.

In work unit 1 of the preceding figure, a reply to the exception response request and CD creates an implied sync point. In work unit 2, sending the DR2 response to the RQD2 creates a sync point.