Kubernetes Patterns, 2nd Edition

Chapter 29. Elastic Scale

The Elastic Scale pattern covers application scaling in multiple dimensions: horizontal scaling by adapting the number of Pod replicas, vertical scaling by adapting resource requirements for Pods, and scaling the cluster itself by changing the number of cluster nodes. While all of these actions can be performed manually, in this chapter we explore how Kubernetes can perform scaling based on load automatically.

Problem

Kubernetes automates the orchestration and management of distributed applications composed of a large number of immutable containers by maintaining their declaratively expressed desired state. However, with the seasonal nature of many workloads that often change over time, it is not an easy task to figure out how the desired state should look. Accurately identifying how many resources a container will require and how many replicas a service will need at a given time to meet service-level agreements takes time and effort. Luckily, Kubernetes makes it easy to alter the resources of a container, the desired replicas for a service, or the number of nodes in the cluster. Such changes can happen either manually, or given specific rules, can be performed in a fully automated manner.

Kubernetes not only can preserve a fixed Pod and cluster setup but can also monitor external load and capacity-related events, analyze the current state, and scale itself for the desired performance. This kind of observation is a way for Kubernetes to adapt and gain ...