DEV Community

fabrizio de luca
fabrizio de luca

Posted on

5 n8n Workflow Mistakes That Will Break Your Production Automations

#n8n #automation #webdev #productivity

Most n8n tutorials show you how to build workflows. This one shows you the 5 mistakes that make those workflows fail in production.

I've been running n8n workflows in production for over a year. Here's what breaks.

Mistake 1: No Error Handling

Every workflow that touches an external API will fail eventually. Rate limits, timeouts, server errors — they all happen.

What breaks:

HTTP Request → Process Data → Save to Sheet
Enter fullscreen mode Exit fullscreen mode

If the API returns 429, the workflow crashes and you never know.

Fix:

HTTP Request → Error Trigger → Retry Logic → Telegram Alert
Enter fullscreen mode Exit fullscreen mode

Add an Error Trigger node at the workflow level. Set up a Telegram notification (30 seconds to configure) so you know immediately when something fails. For retryable errors, use a Wait node + loop back to the HTTP Request.

Mistake 2: Hardcoding Credentials in Node Config

New users put API keys directly in HTTP Request nodes. This means:

  • Keys are visible in exported JSON
  • Can't rotate keys without editing every workflow
  • Can't share workflows safely

Fix: Always use n8n Credentials. Create a credential once, reference it everywhere. When you rotate the key, update one place.

Mistake 3: No Deduplication

Webhook triggers and scheduled workflows both have the same problem: they can process the same item twice.

What breaks: Your email list has 500 duplicates because the workflow ran twice during a server restart.

Fix: Use a Google Sheet or SQLite (n8n has a built-in SQLite node) as a "processed IDs" store. Before processing any item, check if its ID exists. If yes, skip.

Get Item → Check ID in Sheet → [exists] Skip | [new] Process + Log ID
Enter fullscreen mode Exit fullscreen mode

Mistake 4: Synchronous Processing of Large Lists

Processing 1000 contacts one by one in a single workflow execution locks up your n8n instance and times out.

Fix: Split into batches. Use the Split In Batches node with a size of 10-50. Add a Wait node between batches if the API has rate limits.

Get All Contacts → Split In Batches (50) → Process → Wait 1s → [loop]
Enter fullscreen mode Exit fullscreen mode

Mistake 5: No Monitoring

Running workflows without monitoring is flying blind.

What I use:

  • Execution log — n8n's built-in logs (keep last 500 executions)
  • Telegram alerts — 1-line message on workflow success with count processed
  • Google Sheets log — append a row after each run with timestamp + items processed

This gives you a dashboard in Google Sheets showing exactly when each workflow ran and how many items it handled.

The Full Production Template

I built these patterns into ready-to-import n8n workflows. Both packs are free (PWYW):

Each workflow includes the error handling, deduplication, and monitoring patterns described above.

What mistake did I miss? Drop it in the comments — I'm collecting these for a more complete guide.

Top comments (0)