관측 가능성과 분산 추적

Welcome! In modern software, especially with cloud-native and microservices, understanding what's happening inside your system is critical. This is where observability comes in.

Observability is like having X-ray vision into your software. It helps you quickly identify and fix issues, understand performance, and make better design decisions.

Why is observability so important today?

• Complex Systems: Microservices mean many small, independent parts interacting, making it hard to see the whole picture.
• Faster Debugging: Quickly find the root cause of problems when things go wrong.
• Performance Insight: Understand bottlenecks and optimize your system's speed.
• Proactive Detection: Spot potential issues before they impact your users.

Observability relies on three main types of data, often called its "pillars":

• Metrics: Aggregated numerical data collected over time (e.g., CPU usage, request count, error rates).
• Logs: Timestamps and messages describing specific events (e.g., an error message, a user login).
• Traces: End-to-end requests showing the flow and timing across multiple services.

Metrics are numerical measurements collected at regular intervals. They provide a high-level, statistical view of your system's health and performance.

You typically use metrics to:

• Monitor trends over time (e.g., increasing load).
• Trigger alerts when thresholds are breached.
• Understand overall system capacity and usage.

Logs are records of discrete events that occur within your application. Each log entry usually includes a timestamp, a message, and context like the source service or user ID.

Modern systems often use structured logging, where logs are formatted (e.g., JSON) to be easily searchable and analyzable by machines.

Try running this simple logging example:

In a microservices architecture, a single user request can travel through many different services. Distributed tracing helps you follow that request's entire journey from start to finish.

It provides a visual map of how a request flows through your system, showing which services it hits and how long each step takes.

A trace represents the complete end-to-end request. It's made up of multiple spans.

• A span is a single operation within a trace (e.g., an API call to another service, a database query).
• Each span has a unique ID, start/end times, and can have parent/child relationships.

Context propagation is key: it ensures trace IDs are passed along with the request as it moves between services.

Imagine a user adding an item to a cart on an e-commerce site:

• Service A (Frontend): Receives request, calls Service B.
• Service B (Cart Service): Adds item, calls Service C (Inventory) to check stock.
• Service C (Inventory Service): Queries a database for item availability.

A distributed trace would show the timing and sequence of these calls, making it easy to see where delays occur or if a service fails.

Distributed tracing offers significant advantages, especially in complex systems:

• Performance Bottlenecks: Quickly identify slow services or database queries within a request flow.
• Root Cause Analysis: Pinpoint the exact service or component that caused an error or latency spike.
• Service Dependency Mapping: Understand how services interact and depend on each other in real-time.
• Latency Optimization: Focus your optimization efforts on the slowest parts of your system.

Let's test your understanding of observability pillars.

Great job! We've explored observability, which gives you deep insight into your system's behavior.

It's built upon three pillars: metrics (aggregated data), logging (event records), and crucially, distributed tracing (end-to-end request flows).

Distributed tracing is especially vital in microservices for debugging, performance optimization, and understanding complex service interactions.