DEV Community

Alfaz Mahmud Rizve
Alfaz Mahmud Rizve

Posted on Edited on

Your Watchtower can't see a dropout: build a heartbeat check that can

Your error handler only fires when a workflow crashes. A workflow that never starts is invisible to it.

I run a Watchtower: one global error-handling workflow, built on n8n's Error Trigger, that catches crashes across my whole instance and pings me with a deep link to the failed run. It is good at its job. The full build is here: https://whoisalfaz.me/blog/n8n-global-error-handling/

It has two blind spots. The 3.0 upgrade makes the first one urgent, and the second one has always been there.

Blind spot 1: the dropout

In n8n 3.0, workflows built on removed nodes — Cron, Interval, Function — register zero active listeners when the container boots. The dashboard still says Published. No execution ever starts, so the Error Trigger never fires. Nothing crashes, so nothing alerts.

This is the warning box on the Watchtower blog itself: a trigger dropout produces no execution, and no execution means no error event. Your error handler is working perfectly. It simply has nothing to catch.

The first sign of trouble is a customer asking where their invoice went.

Blind spot 2: the dead container

The second blind spot is older. If a workflow eats all the RAM — say, a Code node parsing 50,000 rows at once — the container can die at the operating system level. When the orchestrator itself is dead, the Error Trigger might not fire either, because the thing that would have fired it is gone.

The Watchtower blog is direct about this: hosting is part of the error-handling design, not an afterthought. That is why my production instances run with the V8 memory limit raised. But even good hosting does not close the gap — it only makes it rarer.

The check that covers both

A crash alert answers "did something fail?" A dropout needs a different question: "did something that should have run, run?"

That is a heartbeat check. It is a second, separate workflow that does one job: for each production workflow, look at when it last ran, compare that with how often it should run, and alert when the gap passes the baseline.

Build it like this:

  1. One Schedule Trigger, on its own cadence. Every 15 minutes is a sane default. The watchdog runs on a clock, not on the workflows it watches.
  2. Query the executions API. For each production workflow, pull its recent executions — each run's start time and status. The n8n API exposes this; you are reading the record, not guessing from the dashboard.
  3. Compare against a baseline per workflow. An hourly workflow and a weekly workflow have different "overdue" lines. Record the median gap between runs for each workflow before you rely on the check — the same baseline step the 3.0 SOP recommends before upgrading.
  4. Alert with the same shape as the Watchtower. Workflow name, expected interval, last run, and a deep link to its executions page. One message per overdue workflow, not a dump of the whole fleet.

That is the automated version of the manual check from the SOP: note the gap, watch it, and treat a gap that suddenly goes infinite as a silent dropout.

Two design rules

The watchdog lives on a separate instance from the one it watches. If the watched container dies, a monitor running inside it dies too. This is the whole point of blind spot 2. The simple second home is n8n Cloud — it runs the check, your self-hosted instance runs the work, and neither can take the other down: https://whoisalfaz.me/go/n8n

One threshold per workflow, never one global number. A global "alert if nothing ran in an hour" rule will spam you about the weekly report workflow and stay quiet about the hourly one that died 55 minutes ago. Baselines are per workflow or the check is noise.

The outer ring

The heartbeat check covers dropouts and dead workflows. It does not cover a dead instance, because the check itself needs an instance to run on. That is what the outer ring is for: a plain uptime ping on your instance URL, from outside, that tells you the box is down. Host-level, five-minute interval, no n8n involved.

Three layers, each catching what the one below cannot see:

  • Watchtower catches executions that start and fail.
  • Heartbeat check catches workflows that never start.
  • Uptime ping catches an instance that is down entirely.

Build the Watchtower first — it is the fastest win and the SOP is linked above. Then build the heartbeat check before December, while the 3.0 beta is still out. The workflows that look fine after the upgrade are exactly the ones to check first: https://whoisalfaz.me/blog/how-to-upgrade-to-n8n-3-without-breaking-workflows/


Alfaz Mahmud Rizve — GTM & RevOps Architect. I build autonomous revenue systems for scaling agencies and SaaS: SEO & organic growth, high-speed Next.js apps, and self-healing n8n workflows. whoisalfaz.me

Top comments (0)