How would you use Python's logging module effectively instead of print statements for debugging production code?
Use a configured logger with levels, handlers, and structured context instead of print
print writes to stdout and cannot be filtered, routed, or enriched. The logging module gives you a hierarchy of named loggers, level filtering, multiple handlers (console, file, syslog, JSON), and structured formatting that includes timestamp, level, logger name, and message. Use logger = logging.getLogger(name) so the logger name reflects module structure and can be configured per package. Configure handlers once at the entry point; library code should only create loggers and call methods. Use the lazy % formatting in log calls rather than f-strings so that string construction is skipped when the level is disabled. For production services, emit JSON to stdout and let the platform collect it, and include a correlation or request id via LoggerAdapter or a filter.
- 1
Levels: DEBUG for development detail, INFO for lifecycle events, WARNING for recoverable anomalies, ERROR for failed operations, CRITICAL for process-threatening failures.
- 2
Use logging.getLogger(name) in every module; never use the root logger directly in libraries.
- 3
Call logging.basicConfig or dictConfig once at the entry point, not in libraries.
- 4
Use logger.error('...', exc_info=True) or logger.exception(...) inside except to preserve the traceback.
- 5
Prefer lazy logging: logger.info('user %s did %s', user_id, action) over f-strings, since the format is skipped when the level is off.
- 6
Common mistake: calling logging.info at module import time before configuration, which uses the last-resort handler and prints unexpectedly.
- 7
Common mistake: logging inside a hot loop without a level check, which can dominate runtime.
- 8
Version note: dictConfig and structlog-style JSON logging are widely used in 3.2+. The default lastResort handler behavior has been stable for years.
Scenario Questions
0-2 years experience
- 1What is the difference between logging.info and print?
- 2How do you get the module name into your log messages automatically?
2-5 years experience
- 1Your library prints messages when imported. How do you fix it to use logging correctly?
- 2You need DEBUG logs in staging but only INFO in production. How do you configure this without code changes?
5-8 years experience
- 1You need structured JSON logs with a request id across many modules. How do you set this up?
- 2Your service produces huge volumes of logs and log cost is high. How do you reduce volume without losing debugging value?
8+ years experience
- 1Design an observability strategy combining logging, metrics, and tracing that minimizes overlap and keeps costs bounded.
- 2Explain how logging interacts with asyncio, thread pools, and multiprocessing, and how to keep correlation ids consistent.
Follow-up Questions
- Why is logger.info('x %s', y) preferred over an f-string in a hot path?
- How do you attach a request id to every log line in a web service?