Structured Logging in Production: Why You Should Stop Using console.log (2026)

#node #devops #logging #backend

Your production logs are a wall of unstructured text. Grep fails. Debugging takes hours. You need structured logging.

The Problem

User login failed for john@example.com
Error processing payment
Connection timeout after 30s

Which user? Which payment? Which service timed out? Unstructured logs cannot answer these questions.

Structured Logging with Pino

import pino from "pino";
const logger = pino({ level: process.env.LOG_LEVEL || "info" });

// Instead of: console.log("User login failed for " + email)
// Do: logger.warn({ userId, email, reason }, "Login failed")

Output: {"level":40,"time":"2026-03-21T...","userId":"u_123","email":"john@example.com","reason":"invalid_password","msg":"Login failed"}

Request Context Middleware

Every log in a request should include requestId, method, path, userId:

app.use((req, res, next) => {
  req.log = logger.child({ requestId: crypto.randomUUID(), method: req.method, path: req.path });
  next();
});

Now every log in that request automatically includes the requestId. Trace a single request across your entire system.

Log Levels Done Right

error = broken, warn = recoverable, info = business events, debug = dev context

Sensitive Data Redaction

Never log passwords, tokens, credit card numbers, or PII. Use pino redact:

const logger = pino({ redact: ["password", "token", "req.headers.authorization"] });

Shipping Logs

JSON logs pipe directly to Datadog, Elasticsearch, CloudWatch, Grafana Loki. No parsing regex needed.

Summary

Use Pino (fastest Node.js logger)
Log JSON, not strings
Add requestId to every log via child loggers
Use correct log levels
Redact sensitive data
Ship to centralized logging

Part of my Production Backend Patterns series. Follow for more practical backend engineering.

DEV Community