Cloud-Native Microservices With Kubernetes - 2nd Edition

The 3 Pillars of Observability

In a distributed system made up of many microservices, faults or unexpected behaviors can emerge from complex interactions. To understand why a system behaves a certain way, observability depends on collecting and analyzing telemetry data from each service.

This data falls into three main categories known as the pillars of observability: logs, metrics, and traces.

Logs

Logs are structured or unstructured messages produced when specific code paths execute. They record detailed information about events, such as errors, their source, timestamp, and context. In a microservices setup, logs are invaluable for uncovering the details of unknown faults or new behaviors that appear only under certain conditions.

Analyzing log data helps locate the source of an error, determine when it occurred, and understand why it happened. Logs often form the first line of investigation when diagnosing system issues.

Examples:

• An error log indicating a failed database connection.
• A warning log about high memory usage in a service.

Metrics

Metrics represent numerical summaries of system performance over time. They provide a broader view of what’s happening by capturing trends and patterns rather than individual events.

Examples include uptime, request rate, latency, error rate, CPU load, and memory consumption. Metrics are often used to set alerts or trigger automated responses when values drift outside expected thresholds. Because they’re easy to aggregate and compare, metrics help identify performance trends and emerging issues across multiple components.

Examples:

• A sudden spike in error rates may indicate a problem with a specific service.
• An increase in latency could suggest network congestion or resource contention.

Traces

Traces follow the path of a single request as it moves through different services in a distributed system. They reveal how long each step takes and where bottlenecks or failures occur.

Tracing is especially valuable in debugging complex systems, where a single user action may involve dozens of interconnected services. While logs and metrics describe what’s happening and how often, tracing shows where and why