Guide

Silent failures in automations

Why a workflow with no errors can have been broken for days, what causes it, and how to detect it from the expected rhythm.

By Abdellah El Kharraze · · 2 min read

Key takeaways
  • A silent failure is a workflow that no longer produces its result without reporting an error.
  • Common causes: changed webhook, expired authentication, over-strict filter, deactivated workflow.
  • It is detected through rhythm: maximum delay without success and expected volume.
  • An unreachable API is not silence: the two cases must stay distinct.

A silent failure is an automation that no longer produces its result without reporting any error. The workflow is active, its history shows no failure, and yet nothing comes out. It is the most expensive kind of failure, because it is the one you discover last.

Common causes

  • Webhook changed at the source: the form or application that called the workflow has a new URL.
  • Expired authentication: the trigger can no longer query the service, and the platform does not surface it as an execution error.
  • A filter that became too strict: a field changed name or format, and every item is rejected.
  • Deactivated workflow: by hand during a test, or by the platform after errors.
  • Stopped instance: on self-hosted n8n, a server that is off triggers nothing, including the error workflow.

Why it is expensive

It is discovered late, often by the end customer: unprocessed requests, unsent invoices, unsynchronised orders for days. The cost is not the repair, which is often quick. It is the time that passes before anyone notices.

How to detect it

By comparing real activity with the expected rhythm, using two simple rules:

  • Maximum delay without a successful execution. A workflow that runs hourly and has done nothing for three hours is silent.
  • Expected volume per day. A workflow that handles two hundred items a day and now handles twenty has a problem, even without an error.

The rule must stay deterministic and readable. And it must never confuse an unreachable API with a silent workflow: if the monitor could not read the history, it does not know, and should say so.

How to protect yourself

Three levels, from simplest to most complete: a heartbeat on critical workflows, a regular check for deactivated workflows, then external monitoring that applies these rules to the whole fleet. Nokto does this for your n8n and Make workflows. See how it works.

Frequently asked questions

What is the difference between a failing workflow and a silent one?

A failing workflow runs and ends in error: the platform knows. A silent workflow no longer runs at all, or runs without producing anything: the platform reports nothing.

Which silence delay should I choose?

Two to three times the normal interval between two successful executions is a good starting point, adjusted for quiet days.

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