TL;DR: Observability helps engineers understand what is happening inside a system by examining its outputs. Logs, metrics, and traces provide complementary perspectives that help identify, investigate, and resolve problems.
Remember Grafana? In that article, we discussed logs, metrics, and traces. These are commonly known as the three pillars of observability, or O11y.
Each provides a different perspective on a system.
Metrics: Numerical measurements collected over time, such as response time, CPU utilization, memory usage, and error rate. They help engineers identify abnormal behavior, monitor trends, and trigger alerts when predefined conditions are met.
Logs: Records of events that occur within a system. They provide details about what happened, when it happened, and, depending on the available context, why an operation might have failed.
Traces: Represent the journey of a request as it passes through functions, services, and other components. Distributed traces help engineers identify where latency accumulates and which operations contribute to a failure.
How Do They Work Together?
Imagine an application that suddenly becomes slow.
Metrics reveal an increase in response time and error rate. Traces help identify which service or operation is contributing to the delay. Logs provide additional details about the events and errors associated with that request.
Together, these signals help engineers move from detecting a problem to investigating its cause.
Why Does Observability Matter for Platform Engineering?
A platform may run applications across multiple services, containers, clusters, and cloud environments. Without a consistent observability approach, engineers can struggle to understand failures across these boundaries.
A well-designed platform should make observability accessible by providing standardized instrumentation, centralized telemetry collection, dashboards, alerts, and trace correlation. This allows application teams to investigate problems without having to build the entire monitoring infrastructure themselves.
Observability therefore supports more than incident detection. It also helps teams understand system behavior, evaluate reliability, and make informed decisions about performance and capacity.
Back to Certified Cloud Native Platform Engineering Associate Preparation