Workflow monitoring

With dozens of workflows running on schedules and webhooks, the question “did it actually run?” needs a real answer. I built a monitoring layer on top of n8n.

Components

• Status sync: tracks running, success and failed states for all active workflows and writes them to a MySQL log table.
• Live dashboard: a page with polling, run, retry and stop buttons, an action log per user, and a scan for executions that were never tracked.
• Watchdog: detects stalled runs using two thresholds, sends email alerts, and de-duplicates notifications so one stall means one email.
• Failure logger: a shared n8n error workflow that records the failure reason for multi-trigger chains without splitting them into separate workflows.

Lessons

• Monitoring workflows must not clutter n8n with their own execution data, so they log sparingly.
• Failure-only tracking per trigger chain is far more useful than logging every success.

Stack

n8n, PHP, Python (FastAPI), MySQL.

Memory investigation

I also track the n8n container’s memory use over time with a small sampler, to find out why long-running workflows get unstable and what to change.