Running n8n in production
What changes when a workflow tool stops being a toy: versioning, webhook reliability, exposure, and the failures that produce no error at all.
n8n is easy to start and unforgiving to operate. The gap between a workflow that runs when you click it and one you can leave alone for six months is almost entirely about failure modes nobody mentions in the tutorials.
Three things account for most of the pain.
Edits that do not take effect. The copy the editor shows you and the copy that executes are not always the same thing. Change the wrong one and it applies cleanly, reads back correctly, and never runs. There is no error, which is what makes it expensive.
Webhooks that quietly stop existing. Registration is separate from the workflow itself, so operations that replace the workflow take the registration with them. The workflow still shows as active. The URL returns 404. Nothing is logged.
Exposure. The editor is a full administrative interface over everything the instance can reach, and the webhook endpoints must be public. Treating those as one thing is the most common serious mistake in a self-hosted setup.
Underneath all three is the same lesson: in an automation platform, the dangerous failures are silent by construction. Nothing throws, nothing alerts, and the only symptom is that something you stopped thinking about stopped happening.
Need help with any of this?
These notes are free and always will be. If you would rather someone just set it up, or you are stuck on something similar, get in touch at hello@opsira.io.