Structured Logging

Modern logging practices for production systems. Stop grep-ing, start querying.

The Problem

Traditional logging is broken:

• 17 log lines for a single request = noise, not signal
• String search can't correlate events across services
• Low context = hours debugging at 2am
• Optimized for writing, not querying

Example bad log:

2024-12-20T03:14:24.312Z ERROR Database connection pool exhausted active_connections=20

What user? What request? What service? What's the impact?