Production Patterns and Governance

The Production Mindset for Automation

Automation in production is infrastructure. It has uptime requirements, change management needs, and operational overhead. Treating it like a demo project causes incidents.

Version Control for Workflows

Export n8n workflows as JSON and store in version control. Every change should be:

1. Made in a dev environment
2. Tested with representative data
3. Committed with a descriptive message
4. Reviewed by a second person for high-risk workflows
5. Deployed to staging
6. Monitored for 24 hours before full production rollout

Change Management

Document every workflow: what it does, what triggers it, what it affects, and who owns it. Without documentation, no one knows what will break when you change something.

Access Control

Restrict who can edit production workflows. In n8n: use roles to separate viewers, editors, and admins. Production workflow changes should require at least two approvals.

Observability Stack

Minimum viable observability for production workflows:

• n8n execution log — built-in, shows all executions with status and errors
• External monitoring — ping a health check URL on successful completion; alert if missed
• Cost tracking — log API calls and token usage for AI steps
• Business metrics — track the downstream KPIs the workflow is supposed to move

The Runbook

For every critical workflow, maintain a runbook:

• What does this workflow do?
• How is it triggered?
• What are the common failure modes?
• How do you manually re-process failed items?
• Who to contact in an incident?

Runbooks exist so that anyone can diagnose and fix a workflow incident at 2am, not just the person who built it.