> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ravenna.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Troubleshoot and retry

> Diagnose Ravenna workflows that don't trigger, skip steps, fail, or use the wrong data, then retry runs from the beginning, from the failure, or on the latest version.

When a workflow doesn't do what you expect, the run tells you why. Find the step that went wrong, fix the cause, then retry the run.

## Diagnosis order

1. Find the run (Runs tab, filter by Status or Run ID) and the first failed or skipped step on the path that did not finish.
2. Read that step's failure message, resolved inputs, and runtime log.
3. Classify the cause, then fix it and retry with the right policy.

## Common causes

**Workflow not triggering:**

1. The workflow is not published (switch Off or Paused).
2. Trigger filters exclude the event. Non-matching runs are recorded as Skipped and hidden from run lists, so an empty list points to the filters.
3. An integration the trigger depends on is disconnected.
4. Webhook triggers accept requests only while the workflow is published.

**Step skipped:** an incoming True, False, or approval-outcome connection did not pass, or the step's own filters did not match. Skips cascade down that path. See Conditions and branching.

**Action failing:**

1. A dynamic value resolved to empty or to an unexpected type. Check resolved inputs.
2. The integration is disconnected or lacks the permissions the action needs.
3. The external service is failing.

**Wrong data:** the dynamic value points at the wrong step or field, or the field is absent for this kind of ticket (for example, email details exist only on email-created tickets).

**Slow run:** external calls dominate step runtime. HTTP Request times out at its configured timeout (default 10 seconds, maximum 45). Automatic retries add a few seconds per attempt. Runs stay Running while paused on wait steps.

**Cancelled run:** caused by Stop & Publish, deactivating, or deleting the workflow. Cancelled runs cannot be retried.

## Automatic retry behavior

Each step gets up to 3 attempts, about 3 seconds apart, with a 5-minute limit per attempt. Permanent errors (invalid configuration, missing records) are not retried.

## Retry policies

Retry is available for Failed, Completed with errors, and Completed runs. Every retry is a new run linked to the original as its parent, and it starts from the original trigger data.

| Policy | Version used | Steps rerun |
| - | - | - |
| Retry with latest version | Current published version | All |
| Retry from failure | Original run's version | Failed and not-yet-run steps; completed steps keep their outputs |
| Retry from beginning | Original run's version | All |

* To apply a published fix, retry with the latest version. The other two policies replay the old version.
* Retrying repeats side effects of every step that reruns.
* Via API: `POST /workflows/runs/{runId}/retry` with optional `versionId` and `retryFromFailure`; returns the new run ID. Combining `retryFromFailure` with a newer version reuses the outputs of steps that completed originally, so changes to those steps never execute. Retry from the beginning on the new version instead.

## Skipping waits

Only workspace admins can skip a wait, and only Wait and Wait Until steps that are currently Waiting. Skipping completes the step and continues the run immediately; it cannot be undone.


## Related topics

- [Runs and logs](https://docs.ravenna.ai/documentation/automate/workflows/monitor.md)
- [Workflows](https://docs.ravenna.ai/documentation/automate/workflows/overview.md)
- [Publish and edit](https://docs.ravenna.ai/documentation/automate/workflows/publish.md)
- [Tips & troubleshooting](https://docs.ravenna.ai/guides/how-to/foundry/tips-and-troubleshooting.md)
- [Overview](https://docs.ravenna.ai/documentation/automate/knowledge/overview.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.