Intro to Kubernetes

Working with Kubernetes for Engineers

What Problems does it solve?

Container

A container is simply a process on your machine that has been isolated from all other processes on the host machine.

Another way to think about a container is as a standardized unit of software.

Orchestrator

A piece of software designed to unify many different machines into a cluster, a kind of unified compute substrate, which appears to the user as a single very powerful computer on which containers run on.

Orchestration is coordinating and sequencing different activities. Scheduling is managing the resources available and assigning workloads where they can be most efficiently run. Cluster management joins multiple servers together into a unified, reliable, fault-tolerant (and seamless) group.

Kubernetes

Kubernetes is the (now) standard container orchestration system

Kubernetes does the things that the best system administrator would do: automation, failover, centralised logging, monitoring. It takes what we've learned in the DevOps community and makes it the default, out of the box.

-Kelsey Hightower

Kubernetes Architecture

The Control Plane

kube-apiserver

The front-end server for the control plane

etcd

The database where Kubernetes stores all its information

kube-scheduler

This decides where to run newly created pods

The Control Plane

kube-controller-manager

Responsible for running resource controllers, such as Deployments

cloud-controller-manager

Interacts with the cloud provider (where it exists) managing resources such as load balancers

Worker Nodes

kubelet

Responsible for driving the container runtime to start workloads and monitoring their status

kube-proxy

Performs the networking the routes requests between pods on different nodes and between pods and the Internet

container runtime

Usually Docker but Kubernetes supports other runtimes such as rkt and cri-o

The Application

git clone git@github.com:recipher/ap-docker.git

The sample application has three services - a static web application built with React, a microservice built with Node.js and a database using either sqlite or Postgresql.

We will learn how to Dockerize the application, how to connect the services together, how to ensure data is persisted and how to build a multi-service stack using Docker Compose.

Running the Application

cd ap-docker/app
yarn install
yarn start

Assuming you have node and yarn (or npm) installed, you can run the web application in isolation (the services are faked in development mode).

To access the application, open your browser at http://localhost:3000.

Working with Kubernetes

To run a docker container:

docker run -p 4200:4200 recipher/ap-docker-services

The main tool that engineers will use to work with Kubernetes is called kubectl

kubectl run services --image=recipher/ap-docker-services --port 4200
kubectl port-forward deploy/services 8888:4200

Deployments

A deployment is a supervisor, which continually checks that the container is running and if it ever stops, starts it again immediately

kubectl get deployments
kubectl get deploy
kubectl get deploy -o wide
kubectl describe deploy/services

Pods

A Pod is the Kubernetes object that represents a group of one or more containers (usually one).

Deployments don't manage containers directly because sometimes a set of containers needs to be scheduled together, perhaps sharing storage.

kubectl get pods

Resource Manifests

Using kubectl run to create a deployment is useful but limited. If you need to change something you'd need to delete the existing deployment with kubectl delete and create a new one.

Kubernetes is a declarative system - continuously reconciles actual state with desired state. All you have to do is declare the desired state - the Deployment spec - and Kubernetes will do the rest.

We do this by declaring the spec with a Resource Manifest and use kubectl apply to make the changes.

Resource Manifests

apiVersion: apps/v1
kind: Deployment
metadata:
  name: services
  labels:
    app: ap-docker
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ap-docker
  template:
    metadata:
      labels:
        app: ap-docker
    spec:
      containers:
        - name: services
          image: recipher/ap-docker-services
          ports:
            - containerPort: 4200

Services

A Service resource gives you a single unchanging IP address or DNS name that will be automatically routed to a matching Pod.

You can think of a Service as a web proxy or a load balancer, forwarding requests to a set of backend Pods.

Configuration

The ConfigMap resource is the primary object for storing configuration data in Kubernetes.

A ConfigMap is a set of key-value pairs - once you have a ConfigMap you can supply that data to an application either by creating a file in the Pod or by injecting it into the Pod's environment.

Helm

Resource manifests declare all of the information necessary to deploy your application. But there is a lot of repeated information in those manifests - label selectors, namespaces, ports and so on.

It would be useful to just specify those values once and then reference them whenever they occur in your manifests.

There is a tool for that.

Helm: A Kubernetes Package Manager

Helm

You can use the helm command-line tool to install and configure applications and you can create helm charts, which are Kubernetes packages, which completely specify the resources needed to run the application, its dependencies and its configurable settings.