Failure class

The run is green and nothing happened

The symptom

The execution shows success. No error anywhere. But the record was not created, the message was not sent, and nothing downstream reacted.

What the log shows

node "HTTP Request"  continueOnFail: true
output: [{ "error": "connect ETIMEDOUT" }]
node "Create record"  input: [{ "error": "connect ETIMEDOUT" }]
execution SUCCESS

The question that separates the causes

Is there an error key inside the item payload, rather than at execution level?

Verdict

continueOnFail converts failures into data. It is the correct setting for tolerating individual bad records and the wrong one for tolerating a broken dependency.

The fix

1. Open the failing node → Settings → connect the error output to an explicit branch instead of letting it flow onward.
2. In that branch, alert or write to a dead-letter store — do not silently drop.
3. Turn continueOnFail off entirely on any node whose failure should stop the run.

Hardening

Never combine continueOnFail with a write node unless there is an explicit error branch. Green checkmarks over silent data loss is the worst outcome available — worse than a loud failure, because nobody investigates a success.

Other failure classes

• n8n 400 invalid_grant — and why it comes back every seven days
• Make scenario stopped on 429 Too Many Requests
• Duplicate records that survive rebuilding the Zap
• An import that silently loaded 700 of 1041 records