Runs
Watching a workflow work, approving a step, and finding out what went wrong.
Every execution is a run, kept with its input, its output and what each step did. The Runs tab lists them.
Starting one
Run workflow, with an optional input available to the steps as {{run.input}}. Scheduled and event-driven runs start themselves.
Each run records whether it was Manual or Scheduled, who started it, and which revision of the workflow it used. That last one matters: a run shows what the workflow looked like then, so editing a workflow does not rewrite the history of what already happened.
Run statuses
| Status | Means |
|---|---|
| Pending | Queued, not started |
| Running | In progress |
| Waiting for approval | Paused for a person |
| Completed | Finished |
| Completed with issues | Finished, but some items in a fanned-out step failed |
| Failed | A step errored and the run stopped |
| Cancelled | Someone stopped it |
| Rejected | Somebody rejected an approval step |
| Skipped | A condition skipped over it |
Completed with issues is the one worth looking at rather than filing as success. It means a step running over a list had some items fail while others went through, so the output is real but incomplete.
Approving a step
When a run reaches an approval step, the approvers see Waiting for your approval.
Open the run, read the step, then Approve or Reject, with an optional comment. Approving continues the run from that point. Rejecting ends it.
Comments are worth writing: the decision is kept with the run, so "rejected, the Q3 figure contradicts the dashboard" is still there in a month when somebody asks why that week's digest never went out.
If someone else got there first you are told it was already decided, and if you were not an approver you are told that too.
Reading a run
Open one and you get:
- Input and Output for the whole run
- Timeline of steps with status and timings
- Each step's own output, and its error if it failed
- Items for a step that ran over a list, so you can see which one failed
- Token usage, so you can see what a run costs
Long outputs are truncated on screen and say so.
Cancelling
Cancel run stops one in progress.
Cancelling stops the next step. It cannot undo steps already taken. If the email step has run, the email is sent.
When something goes wrong
A step failed. Open the run and read that step's error. The usual causes are a reference pointing at a step id that does not exist, a tool that is no longer available, or a connection that needs reconnecting.
Empty output from a step. Almost always a reference typo: {{steps.summary.output}} when the step is actually called summarise. Use / to insert references rather than typing them.
"Too many runs are active right now." Your organisation is at its limit for concurrent runs. It will clear.
Nothing ran at all. Check the trigger is enabled and, for event triggers, when it was last checked. A trigger showing "Disconnected" has stopped listening. See Triggers.
It ran but did nothing visible. Look for skipped steps. A condition may be doing its job: on a week with no news, a well-built workflow does nothing.
Deleting a workflow
If runs are in progress, you are warned that deleting cancels them. Run history goes with the workflow.