Troubleshooting a Failing Workflow
How to find why an n8n, Make or Zapier workflow stopped working: the last good run, what changed, the real end state, and when a re-run is safe.
Troubleshooting a failing workflow
This guide is for workflows built on n8n, Make or Zapier that have stopped doing their job. Octacer does not sell a workflow platform. These are the steps our engineers follow when a client's automation fails.
Common symptoms
- Runs show "success" but nothing actually happens.
- A Zap or Make scenario keeps switching itself off.
- A webhook silently stopped firing after a change.
- The CRM and other systems slowly drift out of sync.
- A self-hosted n8n instance keeps going down.
- Customers get confirmations for things that never happened.
Before you change anything
Step 1: Find the last good run
Open the run history and find the last run that did what it should. Everything that changed between that run and the first bad one is a suspect.
Step 2: Ask what changed
Most failures follow a change:
- A login or connection: a password reset, an expired token, a revoked app or a removed user.
- A field, module or app update: a renamed or moved field, a changed form, or a vendor's new module or API version.
- A deployment or upgrade: a new version of the app, or of a self-hosted n8n.
Step 3: Check the real end state
Don't trust the run status alone. Look in the system the workflow writes to: is the record there, with the right values? A run can pass every step and still lose the data. See integration bugs hide behind partial success signals.
Step 4: Read the failing step
Open the failed run, find the first step that failed or returned something unexpected, and compare its input with the input of a good run. The difference is usually the cause.
Common causes and fixes
| Symptom | Likely cause | What to do |
|---|---|---|
| The workflow turns itself off | An authorization error on a connection | Reconnect, then find out why the token expired. See why your automation keeps disabling |
| "Required field missing" | A field was renamed, moved or left empty upstream | Fix the mapping, and add a check that routes incomplete records to a person |
| HTTP 401 or 403 | Expired credentials or reduced permissions | Re-authorize with an account that has the access the workflow needs |
| HTTP 429 | A rate limit | Slow down, batch requests, and retry with increasing waits |
| Duplicates after a retry | No duplicate check before the write | Add a key check before creating records |
| WhatsApp templates rejected with Twilio error 63016 | A template sent outside the 24-hour window without a Content SID | Send the template through a Content SID with variables |
| Every n8n Code-node workflow stops at once | The shared task runner or the instance's memory | See surviving n8n task-runner outages |
Step 5: Change one thing at a time
Make one change, test it on the failing case, and write down the result. If it doesn't help, undo it before you try the next one.
Step 6: Prevent the next one
When to get help
If you can't find the cause, or the failure spans several systems, System Rescue starts with a free 20-minute triage call. One failure in one n8n, Make or Zapier workflow is a $249 Quick Fix: fixed and shown working on the case that failed, or you don't pay. Multi-system problems start with a $199 diagnostic, credited to the repair. Please don't send passwords, API keys or production logs by email; we set up secure access after the call.
Was this article helpful? Thanks for your feedback.
Sorry about that. Tell us what was missing and we will improve it.
Keep reading
Related articles
After-Hours Notification Digest Workflow
Queue after-hours CRM notifications and deliver them in one morning digest without duplicate delivery.
Read the article Best PracticesWhat Is an ERP Approval Workflow? (Roles, Rules, Audit Trail)
An ERP approval workflow is a governed chain—requester, approver, auditor—with rules/matrix routing and a durable audit trail so money, inventory, and compliance decisions leave a portable record, not a chat thread.
Read the article Best PracticesWorkflow Best Practices
Practical rules for workflows that keep running: one owner, validated inputs, failure alerts, retries without duplicates, a run log and documentation.
Read the article
Rather have us set it up for you?
Our engineers build these integrations, workflows, and AI features on your stack — documented, tested, and handed over.