Gather useful ticket evidence
On this page
Evidence should let another person understand what happened without repeating a risky operation or receiving unnecessary private data. Choose specific records and explain why each one matters.
Choose the right evidence#
| Evidence | Useful for |
|---|---|
| API Action Log | Incoming request, route, environment, and overall outcome |
| HTTP Request log | Upstream service request and response failure |
| Workflow run / node history | Execution order, branch selection, node result or error |
| Scheduler activity | Trigger timing, state, and background execution |
| Database snapshot or query result | The relevant record state at a known time |
| Selector information | Configuration scope and environment, with secrets masked |
| Warning | A related operational signal |
| Attachment | Reproduction material, screenshot, or safe supporting file |
Add context to each item#
Record the time with timezone, environment, action/app, and version. Explain the observation: “The request reached FetchOrder and the provider returned 404” is more useful than attaching an unexplained log dump.
Link evidence to the step it supports. A screenshot of an empty table does not establish whether the API returned no rows or the widget used the wrong data path; include the relevant request result as well.
Preserve a useful boundary#
New logs redact recognized credentials and secret-selector values, but ordinary personal information may remain. Review the material before sharing it. Use synthetic records for reproduction and omit tokens, cookies, passwords, and complete environment files.
Historical logs can have different redaction behavior. An erased credential cannot be recovered from a new log export, and a retained snapshot may lack details that were never recorded.
Comment on evidence#
Use evidence comments or reactions to keep the discussion attached to the record it concerns. Summarize conclusions in the ticket rather than leaving the only explanation in a long comment chain.
Avoid collecting a live database snapshot or rerunning a write merely to make the ticket look complete. State the missing evidence and use a controlled reproduction when needed.
Need to investigate a problem? Start with troubleshooting