GDG Processing during Restart

When a job that uses GDG generations is restarted, all references to a GDG relative generation (e.g. +1, -3) will cause the DD statement to reference the same physical file as the original job. The only time that this will not be true is where the processing of the original job (or of other jobs that were run after the original job) has caused the GDG generation to be rolled off, scratched, or deleted.

When a restarted job contains GDG datasets and restart information has been enabled for GDGs, then information about the original high version number and, for GDGs that have a rollover disposition of EMPTY, the empty count is restored when the job is restarted. The JCLCM0945I and JCLCM0946I messages give details of the changes. [1]

In the case of most mainframe jobs there will be no issues but you should be aware of the following scenarios:

1. Where a job runs to successful (non-abend) completion and this results in a GDG generation being rolled off, that generation will not be available on the restart, and any step that references the generation will fail with a dataset not found message.
2. Where a different job, which references the relevant GDGs and creates new relative generations, runs between the end of the job in question and before the job restart takes place, it is possible that the generation referenced by the restarted job will no longer be present.
3. Where a job abends prior to completion, the abnormal termination disposition is applied as specified by the DISP= statements in the JCL. In this scenario you will encounter problems if the step at which you restart references a generation that doesn't exist as a result of the prior abnormal termination.