Structured Logging Patterns — Making Logs Actually Useful

Here's a log line from a real production system:

something went wrong with the order

Quick — is this critical? Is it happening to one user or a thousand? What order? What went wrong? What service generated this? When? Can we reproduce it?

You can't answer a single one of those questions. This log is worse than useless — it's a false signal that someone will waste time investigating.

Now compare:

{"level":"error","message":"Payment processing failed","service":"order-service","orderId":"ORD-7891","userId":"usr-42","error":"Stripe returned 402: card_declined","amount":149.99,"timestamp":"2024-04-15T14:23:03.456Z","traceId":"abc123"}

Same error. But now you know what happened, to whom, how much money was involved, which service, when exactly, and you can trace t