What is Observability?

Monitoring tells you when something is wrong ("error rate is above 5%"). Observability tells you why ("the error spike comes from users in the EU region calling the payment service, which is timing out on the fraud-check microservice due to a slow database query on the orders table"). The distinction matters because modern distributed systems fail in novel ways that predefined dashboards cannot anticipate.

The three pillars work together. Logs provide detailed, event-level records of what happened. Metrics provide aggregated, time-series data showing trends and thresholds. Traces follow a single request as it travels through multiple services, revealing where latency or errors are introduced. Tools like Datadog, Grafana, Jaeger, and OpenTelemetry provide these capabilities.

For QA engineers, observability is a shift-right testing superpower. Instead of only verifying behavior through test assertions, you can verify production behavior through dashboards and alerts. Did the new feature increase p99 latency? Is the error rate for the new API endpoint within SLA? Are there trace anomalies indicating a bottleneck? Observability transforms post-deployment verification from "check that it works" to "understand how it works under real conditions."