Docker?

The premise of docker

BUILD ONCE

RUN ANYWHERE

VMS vs containers?

How works?

Kubernetes?

Kubernetes

It is an open source system to manage containerised applications across multiple hosts, providing basic mechanisms for deployment, maintenance and

scaling of applications

Strengths of kubernetes

• Velocity
   
• Stateless
   
• Declarative configuration
   
• Self-healing platform
   
• Immutability
   
• Programable

Architecture

K8s is composed by a master and minions (nodes). A master contains the key components and nodes connect to it via an secure API thanks to an agent.

Master

• API server
  It is the frontend control panel to manage kubernetes resources/objects
• Control Manager
  It is in charge of maintain the desired state
• Scheduler
  Select in which node a new container has to run
   
• Etcd
  Is a distributed data storage used by kubernetes to store configuration and state

Nodes

• Docker (or rkt)
  Underlying application container runtime
• Kubelet
  Is the agent which talks with apiserver and performs actions requested and monitor the node containers
• Kube-proxy
  Collaborates to maintain the VXLAN network

Networking

It uses a container networking provider to create a VXLAN

• All containers can communicate with the rest without NAT
• All nodes can communicate with all containers (and vice-versa) without NAT
• Every container (pod) has a unique IP within the cluster
• Developer has not to set up network configuration

Kubectl (CLI)

It is a command line interface for running commands against the cluster(s)

kubectl [command] [TYPE] [NAME] [flags]

DEMO

Pod

The most basic resource 

• Wraps an application container(s)
• A unique IP (network)
• Encapsulate storage
• Represents a single unit of deployment
• Can limit resources

Pod

Lets create a pod object definition

DEMO

 apiVersion: v1

 kind: Pod

 metadata:

    name: single-pod

 spec:

    containers:

    - name: my-awesome-app
      image: 172.17.4.1:5000/my-awesome-app:0.0.0

Pod

Lets create a pod object definition with multiple containers

DEMO

app

echo

  shared-data

Multi container pod example

curl

Replica set

Is a controller that helps to control pods

• Before known as Replica Controller
• Maintain a number of replicas
• Used by deployment 
• Managed by HPA (horizontal pod autoscaler) to scale
  number of replicas

Replica set

DEMO

 apiVersion: v1

 kind: ReplicaSet

 metadata:

    name: my-awesome-app

 spec:

    replicas: 3

    template:

      containers:

      - name: my-awesome-app
        image: 172.17.4.1:5000/my-awesome-app:0.0.2

Deployment

Is a controller one level up 

• Control ReplicaSet resources
• Do rollout of new or update of our containers
• Do rollback of current deployments
• Scale up/down containers

Deployment

The specification is the same as previous example

 apiVersion: v1

 kind: Deployment

 metadata:

    name: my-awesome-frontend-app

 spec:

    replicas: 2

    template:

      containers:

      - name: my-awesome-app
        image: 172.17.4.1:5000/my-awesome-frontend-app:0.0.1

DEMO

Service

Group a set of pods under an IP

• Lets pods communicate between them
• Decouple the service implementation
• Match pods  by labels
• Create the endpoints to route requests
• Can resolve domains thanks to a DNS addon
• There are three types of service
  • ClusterIP
  • NodePort
  • LoadBalancer

Service

 apiVersion: v1

 kind: Service

 metadata:

    name: mongo

 spec:

    selector:

​       db: mongo

    ports:

      - port: 27017

        protocol: TCP

DEMO