Logs preserve emitted events that help reconstruct what happened when a distributed system fails. When interactive debugging is unavailable, they can provide the evidence needed to connect symptoms. section 7.6 section 7.6.3
Treat logging as a technical concern that can be kept separate from business logic:
- Inject a logger through an abstraction so implementations can be swapped without changing business logic. section 4.5.3
- Pass a fake logger in unit tests to check that logging is attempted without writing to a real file. section 4.5.3
- Keep business logic focused by assigning logging its own responsibility; frameworks can also weave logging in automatically. section 4.5.3
Apollo 13: connecting the clues
Sy Liebergot correlated the oxygen-pressure drop with fuel-cell power loss. The matching event patterns revealed a shared cause, rather than two unrelated failures. section 7.6.3




















