Advanced Kubernetes: 1 Core Concepts

Course details

Building and operating Kubernetes clusters is a complex topic of increasing importance in today’s DevOps-driven world. Systems engineers, site reliability engineers, and those who work with IT infrastructure at a deep and programmatic level all need to have a next-level understanding of Kubernetes. In this course, instructor Matt Turner reviews its architecture, starting with an overview of the core components of a Kubernetes cluster and the mechanisms for keeping them available at scale. He reviews the etcd database and shows how to investigate the cluster at a low-level through its API. He then reviews the API request process, from authorization to resource persistence, and explores the controllers and reconciliation loops that drive Kubernetes. The closing chapter covers reconciliation in the dataplane, including managing cloud provider resources with the cloud-controller, and how DNS-based discovery of pod and service IPs works.

Concepts

0. Introduction

01 - Exploring the Kubernetes architecture
02 - What you should know

1. Kubernetes Architecture Overview

03 - The Kubernetes cluster architecture from 40,000 feet
04 - Who interacts with the cluster
05 - Storing resources in the database
06 - Making things happen
07 - Achieving high availability
08 - Scaling with load
09 - Bringing it together - Control plane load

2. Etcd and API server - The Core of the System

10 - Some terminology - Kernels, user space, and distributions
11 - Looking at resources in etcd
12 - Manually querying the Cluster API

3. Kubernetes API Server - The Life of a Request

13 - Validating identity with authentication
14 - Checking permissions with authorization
15 - Intercepting requests with admission controllers
16 - Extending admission control with webhooks
17 - Intelligent persistence and retrieval of resources
18 - Bringing it together - apply, diff, and dry-run

4. Reconciliation - The Engine of a Declarative System

19 - Putting the world right with reconciliation
20 - Everything's a controller
21 - Example controller - Deployment
22 - Who controls controllers The controller-manager
23 - The many other loops of the controller-manager
24 - Controlling pods' placement - The scheduler
25 - Bringing it together - Scheduler mechanics

5. Data-Plane Reconciliation

26 - What's contained in a pod
27 - Controlling pods with the kubelet
28 - kube-proxy - The iptables controller
29 - Weather machine - The cloud-controller-manager
30 - Discovering services with cluster DNS
31 - Bringing it together - The operator pattern

Conclusion

32 - Next steps