The CHKPT macro is coded inline in the problem program. When this macro executes, the operating system writes a checkpoint entry in a checkpoint data set. The entry consists of job step information, such as virtual-storage data areas, data set position, and supervisor control, from the problem program.
After the checkpoint information has been written, control is returned to the instruction following the CHKPT macro.
When an application program takes a checkpoint, the system records information about the status of that program in a checkpoint data set. This information includes the location on disk or tape where the application is currently reading or writing each open data set. If a data set that is open at the time of the checkpoint is moved to another location before the restart, you cannot restart the application from the checkpoint because the location-dependent information recorded by checkpoint/restart is no longer valid.
There are several system functions (for example, DFSMShsm or DFSMSdss) that might automatically move a data set without the owner specifically requesting it. To ensure that all checkpointed data sets remain available for restart, the checkpoint function sets the unmovable attribute for each SMS-managed sequential data set that is open during the checkpoint. An exception is the data set containing the actual recorded checkpoint information (the checkpoint data set), which does not require the unmovable attribute.
You can move checkpointed data sets when you no longer need them to perform a restart. The DFSMShsm and DFSMSdss FORCECP(days) commend allow you to use operations such as migrate, copy, or defrag to move an SMS-managed sequential data set based on a number of days since the last access. DFSMShsm recall and DFSMSdss restore and copy are operations that turn off the unmovable attribute for the target data set.
For information about the CHKPT macro, see z/OS DFSMSdfp Checkpoint/Restart.