Handling long-running jobs and programs

For long-running programs, ensure that you do not have a large number of changes that meet these criteria:

Have accumulated over a period of time
Are exposed to the possibility of a backout in the event of a task or system failure

You can avoid problems by using the commit services provided by RRS to split the program into logically separate sections, thus creating units of recovery. The end of a unit of recovery is initiated by a sync point. If a task fails after a sync point but before the task has been completed, only changes made after the sync point are backed out.

A unit of recovery should be entirely logically independent, not merely with regard to protected resources but also with regard to execution flow.

If an abend occurs during either application or DFSMStvs processing, do not issue further requests for the same unit of recovery. The abend can cause some DFSMStvs control blocks to be invalidated. Similarly, do not try to recover from I/O errors. Because of the risk to data integrity, DFSMStvs does not allow a commit of a unit of recovery after it has received a hard I/O error for one of its VSAM record management requests. Also, it does not allow a commit of a unit of recovery for any type of error occurring during a close if a previous explicit commit was not already done.