Your monitoring dashboard displays green across all microservices. CPU utilization is stable, memory looks fine, and average response times seem acceptable.
Then 3 AM hits. Your phone buzzes with a high-severity alert: failure rates are spiking. You open your log viewer, run a broad query for error logs, and get hit with thousands of unstructured log lines per minute. You spend the next hour trying to piece together which service caused the cascading failure.
This nightmare occurs when backend teams treat metrics, traces, and logs as interchangeable tools. They are not. Each pillar answers a fundamentally different question during an incident.
The Three Questions of Observability
To debug distributed systems quickly, your observability stack must clearly separate responsibilities:
Metrics answer What is happening: Metrics are numerical values aggregated over time intervals (like counters, gauges, and histograms). They tell you that your HTTP 500 rate jumped from 0.01%
Discussion
Break the silence
Take the opportunity to kick things off.