How do I handle integration failures, alert someone, or skip a step while testing?

Short answer: When a step fails, no later steps run, so you can record only success back on the form. To learn about failures, turn on the integration's failure notification. To skip an action while testing, put it behind a condition that can't be met.

Why it happens

• An Update an item step placed after a failing HTTP call never runs, so it can't write "failed" to the form.
• There's no trigger for "integration failed", so you can't build a Slack alert that fires on failure.

How to fix it

To get told about failures:

1. Open the integration's More options > Settings.
2. Turn on Notify all collaborators during an integration failure. See integration settings.
3. To see why a run failed, open it in Run history. Click AI Troubleshoot for a suggested cause and fix.
4. Fix the data if needed with Edit payload, then click Rerun. Only failed runs can be rerun. See troubleshooting failed integrations.

For temporary failures such as timeouts or rate limits, turn on Auto retry in the integration header. It retries a failed step up to four times within about three minutes. A retry repeats the step, so use it only where a repeat is safe, for example not on a step that sends an email or creates an item.

To skip an action while testing, for example an email or an HTTP call to production:

1. Add a condition to the action that can never be true.
2. Test the rest of the integration.
3. Remove the condition when you're done.