Production Kubernetes

Chapter 9. Observability

The ability to observe any software system is critical. If you cannot examine the condition of your running applications, you cannot effectively manage them. And that is what we are addressing with observability: the various mechanisms and systems we use to understand the condition of running software that we are responsible for. We should acknowledge that we’re not adhering to the control theory definition of observability in this context. We chose to use this term simply because it has become popular and we want people to readily understand what we’re getting at.

The components of observability can be broken into three categories:

Logging

Aggregating and storing the logged event messages written by programs

Metrics

Collecting time series data, making it available in dashboards, and alerting upon it

Tracing

Capturing data for requests that traverse multiple distinct workloads in the cluster

In this chapter, we will cover how to implement effective observability in Kubernetes-based platforms so that you can safely manage a platform and the workloads it hosts in production. First, we will explore logging and examine the systems for aggregating logs and forwarding them to your company’s logging backend. Next, we’ll cover how to collect metrics, how to visualize that data, and how to alert upon it. Lastly, we’ll cover tracing requests through distributed systems so as to better understand what’s happening when applications are composed of distinct ...