Logging That Operators Read

print goes to stdout and dies there. The logging module routes events to handlers, formats them uniformly, and carries levels for severity:

import logging

log = logging.getLogger(__name__)          # one logger per module

log.debug("cache size: %d", len(cache))    # detail for developers
log.info("imported %d rows", count)        # normal milestones
log.warning("retrying request (%d/3)", attempt)
log.error("failed to save %s", task_id, exc_info=True)   # + traceback

Professional habits baked into that snippet:

• getLogger(__name__) names loggers after modules (app.storage), so operators can silence or amplify each area independently.
• Lazy % formatting, not f-strings — the string is only built if the level is enabled.
• exc_info=True attaches the traceback to error logs; losing it is how "it worked locally" mysteries are born.

Levels are a contract

DEBUG → INFO → WARNING → ERROR → CRITICAL. Choose by who must react: developers tune DEBUG out in production; INFO is the operational heartbeat; WARNING means degraded-but-working; ERROR means an operation failed. Emitting everything as INFO is how logs become unreadable.

Configuration belongs to the entry point

Libraries only get loggers; the application configures them (logging.basicConfig(level=..., format=...)) in one place — usually __main__. That split is why a library can never spam a host app's console.