마이크로서비스의 관측 가능성

Welcome to observing microservices! Microservices break down large applications into smaller, independent services. This brings many benefits, but also unique challenges for understanding system behavior.

Instead of one big application, you now have many small ones, all talking to each other. How do you keep track?

In a microservices architecture, a single user request might travel through dozens of different services, each running on its own server or container. This distributed nature creates several observability hurdles:

• Lost in Translation: It's hard to follow a request's journey end-to-end.
• Debugging Nightmare: Pinpointing the exact service causing an issue becomes complex.
• Inter-service Communication: Understanding how services interact and depend on each other is crucial.

Each microservice generates its own logs. Relying on local log files for each service is impractical. You need a centralized logging solution to aggregate all logs in one place.

This allows you to search, filter, and analyze logs across your entire system, making it easier to see what's happening at a glance and correlate events.

Beyond host-level metrics (like CPU or memory), it's vital to collect service-level metrics. These tell you about the health and performance of individual services.

• Request Rate: How many requests a service handles per second.
• Error Rate: The percentage of requests resulting in errors.
• Latency: How long a service takes to respond to requests.

These are often called 'Golden Signals' for a reason!

Distributed tracing is perhaps the most powerful tool for microservices. It allows you to visualize the entire path of a single request as it hops between services.

Each 'hop' is called a span, and a collection of related spans forms a trace. This creates a clear timeline, showing exactly which services were involved and how long each step took.

How does distributed tracing work across different services? Through context propagation. This means passing unique identifiers (like trace and span IDs) from one service to the next as a request travels.

These IDs are typically included in HTTP headers or other communication protocols. When a service receives a request, it extracts these IDs and uses them to link its own operations to the ongoing trace.

A service mesh (like Istio or Linkerd) can significantly simplify microservices observability. It operates at the network level and can automatically handle:

• Context Propagation: Injecting trace headers without code changes.
• Metric Collection: Gathering request rates, latencies, and error rates for all service-to-service communication.
• Traffic Management: Providing insights into traffic flow and dependencies.

In a microservices world, your service often relies on many other services. If a dependency goes down or slows down, your service might also be affected.

It's crucial to monitor the health and performance of these downstream dependencies. This helps you understand cascading failures and identify the root cause faster when issues arise.

Effective microservices observability isn't about using just one tool. It's about combining logs, metrics, and traces to get a holistic, unified view of your system.

When an alert fires from your metrics, you should be able to jump to the relevant logs and traces to quickly diagnose and resolve the problem.

Which of the following is NOT a primary challenge when observing microservices?

Microservices bring complexity but also powerful observability solutions. We learned about:

• The challenges of distributed systems.
• The importance of centralized logs, service-level metrics, and distributed tracing.
• How context propagation links traces across services.
• The role of service meshes in automating observability.

By combining these pillars, you can gain deep insights into your microservices architecture!