An extract job was canceled in the early hours. Yet the processing job set to start after it with After job ran as if nothing had happened, and SM37 shows it as Finished. If you expected the next job to stop when the first one fails, that looks wrong. After reading this note you can find the checkbox in the After job setting that stops the next job on a day the first one is canceled, see why a chain with that box clear can keep running, and choose the order in which to relink a broken chain.
The basis is the background processing documentation in the SAP Help Portal for ABAP Platform 2025 FPS02 (October 2026), and the SAP ERP 6.0 EHP8 and SAP Library 4.6C help pages that name the checkbox. The job names, dates and times are all invented, and nothing here was run on an SAP system.
"When it ends" and "when it succeeds" are different conditions
While you define a job in SM36, the Start condition button opens the Start Time screen, where the buttons at the top offer start conditions such as Immediate, Date/Time, After job, After event and At operation mode. There is also a separate >> button for starting on a given workday. After job starts the job once the named job ends, and whether that earlier job must end successfully can be set separately. So there are two conditions, and the success condition applies only when you select a checkbox.
With After job chosen, you enter the predecessor in the Name field; with the Start status dependent checkbox selected, the job starts only if the predecessor completed successfully, so it does not start on a day the predecessor is canceled. Which release each of these names and statements comes from is gathered in the last section.
The status table points the same way. Finished means all of the job's steps succeeded; Canceled means the job ended abnormally. That covers a job an administrator canceled in SM37 and a step that failed with an error, for example an E or A message from an ABAP program. Being canceled is one way for a job to end. This note therefore reads it this way: a successor with the checkbox clear starts even when the earlier job is canceled. No single sentence stating this case was found; it follows from two points read together: After job starts the job when the named job ends, and with the checkbox selected the job does not start if the earlier one was canceled.
Jobs scheduled with After job, together with those scheduled with After event and At operation mode, are handled by the event-driven scheduler, and "job completed" is one example of a background processing event. That reads as: the earlier job ending is itself what wakes the next one. To make the next job care how the earlier one ended, you have to select the checkbox. And whatever the condition, the next job starts only when its start condition is met and it has been released.
A fictional example: the extract was canceled, the processing job is Finished
Say the fictional job Z_FI_EXTRACT extracts accounting data and Z_FI_POST reads and processes that data. Z_FI_POST is set to start after the 2026-11-16 run of Z_FI_EXTRACT, with the checkbox clear. The link is to that one run; whether it carries over to the next day is the subject of the fourth section. Z_FI_EXTRACT started at 01:00:00 and ended Canceled 2,460 seconds later, around 01:41:00. Z_FI_POST started three seconds later at 01:41:03, ran for 600 seconds and ended Finished.
Do not take this Finished as a sign that all is well. Finished only says that all of Z_FI_POST's steps succeeded, not that the data it read was complete. If the earlier job stopped halfway, the next job may have processed half-built data. When you see a list like this, check two things first: whether the checkbox in the successor's start condition really was clear, and, in the successor's job log and results, what it read and what it changed. How to put it right depends on what the program does, so this note cannot answer that.
If the checkbox was selected: the order for relinking a stopped chain
With the checkbox selected, Z_FI_POST does not start that night. The documentation read for this note does not say what state Z_FI_POST is left in on SM37 after that. On a day the earlier job is canceled, it is safer to find the successor and look at its status yourself.
When you relink the chain after fixing the cause, one rule applies. When the successor is scheduled, the predecessor must have the status Scheduled or Released. A job that is already running or has already ended does not meet that condition, so it cannot be named as a predecessor. So you cannot attach a new successor to the canceled Z_FI_EXTRACT. The order is: schedule the predecessor again and leave it in a state before it starts, schedule the successor with that new job as its predecessor, then check that both are released. If the Z_FI_POST you set up earlier is still there, delete or reschedule it so that two jobs doing the same work do not remain. That last step is this note's recommendation.
Deleting a predecessor runs into the same link. If you delete a job another job is waiting for, that dependent job can no longer be started, and the system tells you about such successor jobs. You then need to reschedule or delete the successor. If you try to release a successor whose predecessor was deleted, the system sets it to Planned. In the status table, Planned means the job's steps are defined but there is no start condition yet, so to run that job you have to specify a start condition again and release it.
This Planned appears to be the same status as the Scheduled in the relinking rule above. The eight statuses in the status table include no Scheduled, and the job chain description and the JOB_CLOSE sample give the predecessor condition as Scheduled or Released. For the same SM37 action of revoking a job's release, the page on releasing jobs says the status goes back to Planned, while the job overview description names the menu item for it Scheduled and, under Check Status, also says a job status can be reset to Scheduled. This note reads it as one status under different names in different pages; which label SM37 shows on screen was not checked.
A successor cannot be a periodic job
A job scheduled to start after an earlier job completes cannot be scheduled as a periodic job. In the first figure, that means you cannot keep After job and also select the Periodic job box. Say Z_FI_EXTRACT is a periodic job that runs every day at 01:00 and Z_FI_POST is attached with After job. Z_FI_POST itself is not a periodic job, so it does not repeat on its own. Whether it is attached again to the new run of Z_FI_EXTRACT on the second day could not be confirmed in the documentation read for this note. Because a program names the predecessor by its job name (PRED_JOBNAME) and job count (PRED_JOBCOUNT) together, this note reads the link as tied to one particular job and not attached again to a new run. If you have set things up this way, check in SM37 on the second day whether Z_FI_POST is waiting again.
The job chain example in the SAP help covers a chain that has to be scheduled irregularly, not on a fixed timetable. There, a single first job schedules the whole chain: both the second and the third job. No job can set up a successor for itself, because when a successor is scheduled, the predecessor must be Scheduled or Released, and a running job does not meet that condition. The example shows two relationships. Between the first and second job there is no formal link, a loose relationship: the second usually runs after the first, but the two can overlap. Between the second and third job the predecessor mechanism makes a tight relationship: the second job is guaranteed to end first, and with PREDJOB_CHECKSTAT in JOB_CLOSE as well, the third job starts only if the second succeeded. It means the same as the checkbox on the screen.
The same example names an alternative: making each job of the chain a periodic job. It says this suits only jobs that have to follow a fixed timetable, and warns that nothing schedules the jobs again automatically when one of them fails. The page on periodic jobs adds that a periodic series keeps going even when one of its runs ends abnormally, though a run that never executes at all gets no follow-on run. This note reads the two together: the failed run is not repeated, and the other jobs still start at their own times without waiting for its result, so a later job can run on data an earlier job left incomplete. For a chain that has to run every day, you could make the first job that schedules the whole chain a periodic job, but that combination is not described in the pages read for this note; it is this note's reading.
Releases and scope
That you can choose whether the earlier job must succeed is in the ABAP Platform 2025 documentation, but that documentation does not name the checkbox. Start status dependent and Name are how the SAP ERP 6.0 EHP8 help writes them. The SAP Library 4.6C help, a page exported in 2001, shortens the same checkbox to Start status depend., and among the pages read, only this 4.6C help says the job does not start if the earlier one was canceled. The label on today's screen was not checked, and releases other than these three were not compared. Job chains built in the Fiori app Application Jobs are not covered in this note.
When you set up After job, ask one question: may this job run on a day the earlier job fails? If not, select the checkbox, and on a day the earlier job is canceled, find the successor in SM37 and attach it to a new predecessor. If it may, leave the box clear, but remember that the next job's Finished does not mean the earlier job succeeded.
Add your perspective.
Share a question, another approach, or something you have tried.
Checking sign-in…
Loading comments…