Add a parameter annotated context: dg.AssetExecutionContext and Dagster passes the execution context: the run id, the partition, the resources and a logger. Messages sent with context.log.info(...), .warning(...) or .error(...) become events in the run's event log, with a time, a level and the step that wrote them. You can filter them in the run page and alerting tools can read them.
print() still works, but it goes to the step's captured output, not to the structured event log. Prefer context.log for anything a person should find later.
Log facts that help the next person: how many rows came in, which day was processed, why something was skipped. Use warning for conditions that need attention but should not stop the run, such as pending orders that will be picked up tomorrow.
The message appears as a LOG event in the run page, on the raw_orders step.