In Chapter 3, Getting Started with Kubernetes, we introduced the basic functions of Kubernetes. Once you start to deploy containers with Kubernetes, you'll need to consider the application's data life cycle and CPU/memory resource management.
In this chapter, we will discuss the following topics:
Kubernetes and Docker use a local disk by default. The Docker application may store and load any data onto the disk, for example, log data, temporary files, and application data. As long as the host has enough space and the application has the necessary permission, the data will exist as long as a container exists. In other words, when a container terminates, exits, crashes, or is reassigned to another host, the data will be lost.
In order to understand Kubernetes' volume management, you'll need to understand the Docker volume life cycle. The following example is how Docker behaves with a volume when a container restarts:
//run CentOS Container
$ docker run...
In general, the application is designed to be long-lived in the same way as the daemon process. A typical long-lived application opens the network port and keeps it running. It is required to keep running the application. If it fails, you will need to restart to recover the state. Therefore, using Kubernetes deployment is the best option for the long-lived application.
On the other hand, some applications are designed to be short-lived, such as a command script. It is expected to exit the application even if it is successful in order to finish the tasks. Therefore, a Kubernetes deployment is not the right choice, because a deployment tries to keep the process running.
No worries; Kubernetes also supports short-lived applications. You can submit a container as a Job or Scheduled Job, and Kubernetes will dispatch it to an appropriate node and execute...
In this chapter, we covered stateful applications that use persistent volumes. Compared to ephemeral volumes, they have some pitfalls when an application restarts or a pod scales. In addition, persistent volume management on Kubernetes has been enhanced to make it easier, as you can see from tools such as StatefulSet and dynamic provisioning.
Furthermore, Jobs and CronJobs are special utilities for pods. Compared to deployment/ReplicaSets, this has a desired number of pods running, which is against Job's ideal situation in which the pods should be deleted once they finish their tasks. This kind of short-lived application can be managed by Kubernetes as well.
In Chapter 5, Cluster Administration and Extension, we will discuss the cluster administration such as authentication, authorization, and admission control. We will also introduce the Custom Resource Definition...
© 2019 Packt Publishing Limited All Rights Reserved
