Running Stateful Applications with Kubernetes

In this chapter, we will learn how to run stateful applications on Kubernetes. Kubernetes takes a lot of work out of our hands by automatically starting and restarting pods across the cluster nodes as needed, based on complex requirements and configurations such as namespaces, limits, and quotas. But when pods run storage-aware software, such as databases and queues, relocating a pod can cause the system to break. First, we'll explore the essence of stateful pods and why they are much more complicated to manage in Kubernetes. We will look at a few ways to manage the complexity, such as shared environment variables and DNS records. In some situations, a redundant in-memory state, a DaemonSet, or persistent storage claims can do the trick. The main solution that Kubernetes promotes for state-aware pods is the StatefulSet (previously called PetSet) resource, which allows us to manage an indexed collection of pods with stable properties...

Stateful versus stateless applications in Kubernetes

A stateless Kubernetes application is an application that doesn't manage its state in the Kubernetes cluster. All of the state is stored outside the cluster and the cluster containers access it in some manner. In this section, we'll learn why state management is critical to the design of a distributed system and the benefits of managing state within the Kubernetes cluster.

Understanding the nature of distributed data-intensive apps

Let's start from the basics here. Distributed applications are a collection of processes that run on multiple machines, process inputs, manipulate data, expose APIs, and possibly have other side effects. Each process is a combination of its program, its runtime environment, and its inputs and outputs. The programs you write at school get their input as command-line arguments, maybe they read a file or access a database, and then write their results to the screen or a file or a database...