Troubleshoot a workflow
Match what you are seeing to the right fix: failed runs, paused runs, broken connections, publish errors, and triggers that go quiet.
Start with the symptom
Pick the row that matches what you are seeing and follow the link. Almost every problem falls into one of three buckets: a run that failed, a run that paused or is waiting, or a workflow that looks live but is not receiving events. If you are not sure which status you are looking at, the run and step status reference defines every value and the exact next action for each one.
| What you see | Likely cause | Where to go |
|---|---|---|
A run shows status failed | A step errored and the run stopped | Read a failed run |
A run shows status paused or waiting | An operator paused it, or it is parked on an approval, a reauth, or a spend cap | Why a run paused |
| A red banner says a connection needs to be reauthorized | A connection's token can no longer be refreshed | Reconnect an integration |
| You fixed the problem and want to run it again | Replay a terminal run as-is | Re-run a run after you fix it |
| A toast says "Workflow compilation failed", or publishing is blocked | A validation rule is failing the publish gate | Fix errors that block publishing |
| The workflow looks live but nothing comes in | Not published, version not active, or the trigger never fired | Fix a trigger that is not firing |
The run status paused means one run stopped, and you resume it with the Resume button on the
run detail page. A workflow status of "Paused (safety)" means the whole workflow stopped
dispatching because a safety breaker tripped, for example a throughput or spend-cap problem, and
you resume it from the workflow page. They surface in different places and resume differently.
Why a run paused covers both.
Two checks before you read deep
Before you trace a single run end to end, two checks resolve a large share of "it stopped working" reports:
- Is the workflow published and its version active? The status badge should read "Published" with the tooltip "Active and dispatching.", and the live version's publish toggle should show Active. A
draft,inactive, ordisabledversion receives no live events even while the workflow page is open in front of you. Full walkthrough: Fix a trigger that is not firing. - Does the notifications bell show a reauth alert? A connection that can no longer refresh its token raises an
oauth-reauth-requirednotification and a red reauthorization banner. A single transient auth blip does not raise this and does not need a reconnect, because the run retries on its own. Only a permanent refresh failure surfaces the banner. When it does, reconnect the integration.
What these pages do not cover
These pages help you read the statuses, errors, and banners that TaskJuice surfaces, and recover from them yourself. A few stops are platform guardrails rather than something you resume by hand:
- Hard spend-cap pauses resume on their own once the cap is raised or the billing period rolls over. There is no manual "Resume workflow" button for them. The banner offers an "Adjust cap" link to billing instead. See Why a run paused.
- Platform abuse limits reject a step rather than pausing a run you can resume.
- Compiled-plan integrity errors, with messages like "Missing or invalid entryStepId", indicate an internal fault, not a misconfiguration you can fix in the editor. If you ever see one, contact support.
Related
- Run and step statuses explained defines every status value and its next action.
- Read a failed run walks the error code, message, and provider status step by step.
- Fix a trigger that is not firing covers triggers that go quiet.