Dynamic Backout

Dynamic backout undoes the changes of a single failed task while CICS keeps running. It is the everyday recovery mechanism - most task abends are cleaned up this way with no operator action.

What dynamic transaction backout does

• When a task abends, CICS automatically backs out all changes the task made to recoverable resources in its current unit of work.
• The rest of the region is unaffected - other users keep working while the failed task's updates disappear.
• This is called dynamic transaction backout (DTB) because it happens dynamically, without restarting anything.
• DTB also runs when a task is purged (for example by CEMT SET TASK PURGE) or when the program issues SYNCPOINT ROLLBACK.

Limits of dynamic backout

• Only recoverable resources are backed out. Updates to non-recoverable files, unprotected TS queues and terminal output stay as they are.
• Records added to a VSAM ESDS during the unit of work cannot be physically removed by backout - they are marked as deleted instead.
• If backout itself fails, the unit of work is marked BACKOUTFAILED and the record must be repaired manually.
• Backout uses the journal, so if journaling was not active for the resource, backout is impossible.

Forcing a backout from the program

• A program can deliberately undo its unit of work with SYNCPOINT ROLLBACK, typically in an error paragraph:-
  9000-ERROR-EXIT.
  EXEC CICS
  SYNCPOINT ROLLBACK
  END-EXEC.
  MOVE 'Update failed - no changes were saved.' TO MSGO.
  EXEC CICS
  SEND MAP('ERRMAP') MAPSET('ERRSET') ERASE
  END-EXEC.
  EXEC CICS RETURN END-EXEC.
• After ROLLBACK the program can continue with a new unit of work, retry the operation, or end the task.
• Use this pattern when validation fails after some updates have already been written.