Capture the business symptom
Record the missing, duplicated, late, or incorrect outcome and the approximate execution time.
Home / n8n troubleshooting
Failure field guide / n8nThe workflow ran. Did the business outcome happen?
A green execution is not the same as a dependable process. Start with the failed business outcome, then trace the exact execution, input, transformation, external call, and recovery path.
- ObserveExecution and node evidence
- ReproduceSame input, controlled run
- IsolateFirst incorrect boundary
- RecoverRetry, review, or stop
Diagnostic order
Trace evidence before editing nodes.
Find the matching execution
Compare the trigger payload, node inputs and outputs, timestamps, and the last trustworthy step. n8n’s execution history is the primary starting point.
Separate platform failure from logic failure
An explicit node error, a successful run with wrong data, and a run that never started require different fixes.
Test recovery safely
Before retrying, check whether earlier steps already created records, sent messages, or changed state. A blind replay can duplicate work.
Common failure classes
What to inspect when “it sometimes breaks.”
- Unexpected input shape or missing required fields
- Expired credentials, permissions, or environment changes
- Rate limits, timeouts, and temporary upstream API failures
- Partial completion followed by an unsafe whole-workflow replay
- No error workflow, alert, or owner handoff
- Successful technical execution but incorrect business logic
Do not paste credentials, API keys, customer records, or sensitive production data into a diagnostic request. Exports, representative samples, logs, or read-only access are safer starting points.
Rescue decision
Repair the boundary—not the screenshot.
A responsible repair identifies the first incorrect boundary, the affected records, the safe replay point, and the monitoring needed afterward. If the workflow cannot be explained or recovered without unacceptable risk, a bounded rebuild may be safer.
This page is educational. Augmentive AI’s paid diagnostic separately scopes one existing workflow, available evidence, and the access required to reach an evidence-backed rescue-or-rebuild recommendation.
Primary source
Have an n8n failure you can show?
Describe the symptom and available evidence. The first step is a fit review, not an automatic production change.
Request fit review →