Skip to main content

Execution

A route works in individual passes, and each of them is a run — with its own time and its own outcome. This page answers what a run creates and why its outcome is something other than the outcome of a single record. When a run starts is described under Triggering.

The run and the jobs in it​

What a run takes along breaks down into jobs: a job is a single record on its way. The run bundles the jobs it created — and has an outcome of its own besides.

A run can end successfully although some of its jobs turned out differently. That is why, next to the outcome of a run, Xchange always also states how its jobs turned out.

In the example: The run of "Statusmeldungen Kessler" at 9:05 AM created 17 jobs. Its outcome is Succeeded, and the Jobs column says "17 jobs: 1 for review, 16 succeeded". Both are true: 16 messages have arrived on the other side, one is waiting for a decision, and the pass itself came to a smooth end.

How the outcome of a run comes about​

At the end of a run, each of its jobs falls into one of three classes:

ClassWhat belongs in it
Had an effecttransferred — or put to a person for review
Failedfailed permanently or stuck behind a failed job
Nothing to doskipped, for one of the reasons in the next section

The outcome of the run follows from this. If the class Failed stays empty, the run is Succeeded. If something failed, what decides is whether something had an effect besides: then it is Partially succeeded, otherwise Failed. Only the first two classes count for the outcome; how many jobs were skipped is shown in the breakdown next to it.

Two consequences of this:

  • A run that reaches into nothing is Succeeded. It looked, and the list was empty; the Jobs column reads "No jobs". For a route on a short cycle, this is the most frequent run of the day.
  • A run whose jobs were all skipped is Succeeded as well. In the example: "Artikel Webshop" ran at 8:31 AM in response to an event — an item had been changed, but the changed fields all lie outside this route. The Jobs column reads "1 job: 1 skipped". If, on the other hand, only failures stand next to skipped jobs, the run is Failed.

Which outcomes there are in total and what each is called in detail is listed in Execution.

Why a job is skipped​

The single job shows Skipped with an addition in parentheses, and the addition says what caused it:

AdditionWhat is behind it
no changeThe mapping yields the same as at the last transfer, and the record has not changed in the target system since.
staleA newer state already exists for the same record; the older one is dropped.
echo detectedThe change has already passed through Xchange via the same object of the target system this route writes to, and would only go back there, so it ends here. If it travels on through SAP Business One to a system or object that does not have it yet, it is transferred.
conditionA condition in the flow has excluded the step for this record.
deletedThe source has flagged the record as deleted, and the route's delete behavior is Do nothing.
never transferredThe source has flagged the record as deleted, but the route never transferred it to the target system — there is nothing to deactivate there.
not presentA step is meant to change a record with Update that does not exist in the target system. There is nothing to change, and open clarification cases for the record close.
clarification openThe same record is already waiting in clarification; this run therefore leaves it aside.

One of these eight stands out. The addition clarification open carries the same word Skipped, but counts as a failure for the run outcome — the job is stuck behind an open case. In the breakdown below the run outcome it is therefore called blocked.

These reasons apply to runs that go through a dataset. A targeted run — from Transfer individually or from a retry in clarification — transfers what is named: without comparison against the state last transferred, and even when a case is open for the same record. That is what the two buttons are for. More on this under Push manually.

Temporary or permanent — this decides who acts​

If a job fails, the kind of error decides what happens next:

  • Temporary. The target system was unreachable at the moment, took too long or was overloaded. The next attempt may succeed, so Xchange retries on its own. You can wait.
  • Permanent. The answer on the tenth attempt will be the same as on the first — a mandatory field is empty, a key cannot be matched, the target system rejects the content. Retrying is pointless, and the job goes to Clarification.

Whether a record from clarification runs along again in the next run depends on what kind of case is there. An open failure holds it back: the next run that picks it up again leaves it aside, instead of letting it fail a second time and filling the list. An open review lets it keep running — it is an open question, without a failure that could be repeated.

The holding back of a failure can end, too: if the record changes in the source system, it is a new state instead of the same failure, and the next run takes it along. What it depends on in detail is described under Clarification.

Only a certain kind of route can pick it up again, however. If a route reads its dataset repeatedly, it comes across the same record in every run — there the rule applies. If, on the other hand, a route takes its work from a queue, the message is used up after the first run; the case then lies in clarification alone, apart from all further runs.

In the breakdown below the run outcome, a permanently failed job is called "awaiting clarification" — after the screen on which it now lies. "cancelled" stands there for something else: for a job that a cancellation stopped.

In the example: The order WS-2026-104861 came from the webshop without a customer number. That stays the same on every retry, so it has been in clarification since the run at 4:20 PM the day before. "Bestellungen Webshop" takes its work from a queue — the runs since then have passed it by.

Several runs of the same route at the same time​

A route may have several runs at the same time — whether two get along depends on their kind. The dividing line runs where a run goes through the entire dataset. The full and the incremental ones do.

Two runs that go through the dataset exclude each other — in any pairing, including two full ones. This already holds while the first one is only requested and still waiting to start.

A run from the queue waits for them. It does not start while a run is going through the dataset. The other way round, such a run may begin while one from the queue is working. Two runs from the queue of the same route exclude each other as well.

Everything else starts at any time. A run triggered by an event in SAP, one from the dispatch input and a targeted run touch named records instead of the entire dataset, so they get past any other run. With events this is the normal case — the run at 9:10 AM can begin while the one at 9:05 AM is still running. And you can push a single record while a large run is still busy.

A refused run is never created, and the running one stays untouched. If you start it by hand, Xchange reports A run for this route is already requested. If a slot of the schedule falls into such a run, that slot is dropped without anything showing in the run list. The schedule continues with its next regular slot, not as soon as the other run ends. Nothing is lost: the next run goes through the dataset again anyway.

Why a cleanup waits instead of reporting​

The Cleanup finds the records that this route transferred earlier and that today fall outside the data filter. It recognizes them through a comparison that only the full run can make: what once arrived on the other side, against what the filter covers now. What lies in the difference is gone from the target system.

A reported removal takes something away in the target system — even when the difference only comes from a data filter that was changed by mistake.

That is why a limit applies: if a run reports an unusually large number of removals, it stops and waits for your decision before it transfers them. While it waits, the target system and the route stay unchanged. You then decide between two ways, and both are final only for this one run: releasing reports the removals, discarding holds them back. After discarding, the difference remains, and the next full run presents it again.

What a run outcome leaves open​

  • Whether the right thing is in the target system in business terms. Xchange reports what the target system answered. An order that was accepted there but assigned wrongly is a successful job.
  • Whether everything was taken along that should have been. That is decided by the data filter of the route — see Routes — Source data.
  • How much ran in a day. A run is a single pass; the daily figures are on the Overview.
  • How much is still waiting at the external system. If a route reads from a queue, a run counts only what it took along itself. What is still lying there is not yet in Xchange — and is therefore also missing from the key figure Backlog.

See also​