Reproduce the Symptom
Before explaining a failure, make it happen on demand and capture the exact inputs, environment, and output.
The question
Can I produce the same failure twice?
Do this
- Separate the visible symptom from a guessed cause.
- Minimise the reproduction without changing the symptom.
- Keep request ids, timestamps, inputs, and versions with the report.
Example
- “Payment is broken” becomes “POST /checkout for cart 1042 returns 402 in production, succeeds in staging, and produces request id req_7d2c41.”
Ask next
- ?What exactly is observable?
- ?Which input or environment is necessary?
- ?What is the smallest reproduction?