Every run of every workflow has a row here, along with what each step took in and gave out.
The statuses are not interchangeable
- Running — not finished yet.
- Waiting — it has reached a delay or a wait-for-event node and is standing by for that to happen. This is not a failure.
- Skipped — rate control meant it never started.
- Completed or failed.
A waiting run can sit for days and then continue by itself; nothing is needed from you.
Two kinds of retry that differ
Rerun runs the workflow from the beginning with the same trigger data. Everything done the first time is done again.
Retry from the failed step takes the successful steps' output from the earlier run and continues from the step that failed. The successful steps are not run again.
For a workflow that sends an email or creates an invoice, the difference between the two is the difference between once and twice.
Retry from the failed step is not always available
It works only for a run whose status is failed and which recorded a snapshot of its graph. Very old runs have no snapshot, and only a rerun is possible for them.
Every retry is a new run
The result is a new row in this list, linked to the earlier one — the old row is not overwritten. So the record of the failure is never erased, and how many attempts were made is always visible.
Steps keep their input and output
For each step you can see what went in and what came out. When a {{steps.…}} expression produces something unexpected, the answer is here.