kimtth Goal: logs that are machine-queryable and useful during incidents.
Use for:
- adding or cleaning up application logging
- making logs searchable and correlatable
- fixing noisy, unstructured, or leaky logs
Workflow:
- Log structured key-value events, not free-form strings.
- Attach context: request ID, user ID, and trace ID.
- Use levels deliberately: error, warn, info, debug.
- Log at boundaries and on state changes, not every line.
- Redact secrets and PII before they are written.
- Verify logs answer "what happened to this request".
Practices:
- one event per significant action, with fields
- consistent field names across services
- correlation IDs to stitch a request together
- sampling for high-volume debug logs
Rules:
- never log secrets, tokens, or PII
- use levels consistently; do not log everything as info
- prefer structured fields over string interpolation
- make every error log actionable, with context