Guide

How to monitor n8n workflows in production

Error workflow, heartbeat, external monitoring through the API: what each method detects, what it misses, and how to combine them.

By Abdellah El Kharraze · · 2 min read

Key takeaways
  • The n8n error workflow only fires when an execution fails.
  • It sees neither the workflow that no longer starts, nor the stopped instance, nor the deactivated workflow.
  • External monitoring through the API catches those cases without changing your workflows.
  • Each workflow needs its own silence delay.

Monitoring n8n in production means answering three questions for every workflow: did it fail, is it still running, and is it producing what it should? n8n answers the first one well. The other two need an outside view.

The three monitoring methods

MethodDetectsDoes not detect
Error workflow (Error Trigger)An execution that failsWorkflow that no longer starts, stopped instance, deactivated workflow
Heartbeat at the end of the workflowThe absence of executionsVolume, duration; must be maintained in every workflow
External monitoring through the APIFailures, deactivations, silence, volume, slowness, unreachable instanceThe business content of the data

The error workflow: useful but blind

n8n lets you attach an Error Workflow to each workflow, in its settings. It starts with an Error Trigger node and runs when a production execution fails. It is the first thing to set up. But it does not fire when the workflow no longer starts, when the instance is down, or when someone deactivates the workflow. And you have to remember to attach it to every new workflow. The details are in this article.

The heartbeat

Adding an HTTP call at the end of a workflow to an external service lets you detect the absence of executions: if the signal stops arriving, something is wrong. The method works, with two caveats. You have to maintain it in every workflow, and it says nothing about volume or duration. See the heartbeat how-to.

Monitoring through the API

The n8n public API exposes the list of workflows, whether each is active, and executions with their status and dates. An external service can query it regularly and detect, without changing a single workflow: failures, deactivations, no successful execution within a delay, a drop in volume, and an instance that stops responding.

What you need to tune

  • A silence delay matched to each workflow's rhythm: one hour for a frequent sync, several days for a monthly task.
  • An expected daily volume for high-traffic workflows.
  • A clear rule for incomplete data: never conclude silence when the API did not answer.
  • A long enough execution retention period in n8n, otherwise the history disappears before it is read.

The right combination

Keep the error workflow for the technical detail of a failure, and add external monitoring for everything it cannot see. That is Nokto's approach: it reads your instance's API every two minutes and works out each workflow's rhythm by itself. Try it free for 14 days.

Frequently asked questions

Does n8n warn you when a workflow stops triggering?

No. n8n reports a failed execution, not the absence of executions. You need a heartbeat or external monitoring.

Does the error workflow run for manual executions?

No. It fires for production executions of an active workflow, not for tests run by hand in the editor.

AE
Abdellah El Kharraze

Founder of Nokto and Owepo, based in Nice, France. Network and security engineer for large enterprises until 2016, then ten years building n8n and Make automations for small businesses.

Your first monitored workflow in ten minutes.

Connect an n8n instance or a Make team: Nokto discovers your workflows, learns their rhythm and warns you at the first anomaly.

14 days free · no credit card · cancel anytime