< 8 Docker on Windows and Mac | 10 Conclusion >

44: Container Orchestration

https://www.udemy.com/course/learn-docker/learn/lecture/15828644#content

• You can only deploy a single container with the docker run command.
• What if the container fails? You’ll need to manually relaunch the container.
• What if the load increases? You’ll need to manually launch additional containers.
• What if the Host fails?  Who will sit and monitor that?

This is what Orchestration solves

Options for Orchestration

• Docker swarm
  • Easy to setup
  • Lacks advanced features
• kubernetes
  • Slightly difficult to set up
  • Provides a lot of options to setup and customize.
• MESOS
  • Difficult to configure

45. Docker Swarm

https://www.udemy.com/course/learn-docker/learn/lecture/7894030#content

Docker swarm allows you to arrange your containers into clusters

• Manage across hosts
• Scalability
• High Availability

Swarm Setup

Create the Swarm manager using

docker swarm init

Join the worker nodes

docker swarm join --tocken <token>

Docker service

A Docker Services are one or more instances of a single application that run across the nodes in a cluster.

On Master

docker run my-web-server
docker service create --replicas=3 <-e> <-p> <--network> my-web-server

46: Kubernetes Introduction

https://www.udemy.com/course/learn-docker/learn/lecture/15828702#content

Kubernetes actually requires it own courses. (5?)

kubectl scale --replicas=2000 my-web-server
kubectl run --replicas=1000 my-web-server

Rolling upgrades / downgrades

kubectl rolling-update my-web-serve --image=web-server:2
kubectl rolling-update mys-web-server --rollback

Testing upgrades

Possible to test upgrades by only rolling out the upgrade to a percentage of instances in your cluster.

Open Architecture

• Almost every vendor you can think of provide plugins for Kubernetes.
  • Storage
  • Networking
• Available on all popular Cloud hosts
• Works with various containerization systems
  • Docker
  • rkt (Rocket)
  • Others

Understanding Kubernetes Architecture

• A Kubernetes cluster consists of a set of nodes
• A Node
  • Is a physical or virtual machine on which the Kubernetes tools are installed
  • Is a worker machine where containers will be launched.
  • For HA, you need to have more than 1 node.
• A Cluster is a set of Nodes grouped together.
• A Master
  • Is a Node with Kubernetes Control Plane Components installed.
  • Watches over the Nodes and is responsible for the orchestration of the containers.

Components

• API Server
  • Acts as the front end for Kubenetes
  • Management devices, CLIs, all communicate with Kubernetes via the API
• ectd
  • Key: Value store
  • Stores all data used to manage the cluster
  • Stores all information about all nodes and masters in a distributed manner.
  • Responsible for implementing logs to ensure there are no conflicts between the masters
• scheduler
  • Distributes the work across multiple nodes
  • Looks for newly created containers and assigns them to nodes
• Controller
  • Brains behind orchestration
  • Responsible for noticing and reacting when Nodes, Containers or Endpoints go down.
  • Controllers make decisions to bring up new containers in such instances.
• Container Runtime
  • The underlying software used to run containers.
    • Docker, etc.
• kubelet
  • Agent that runs on each node in the cluster
  • Makes sure the containers are running on the nodes as expected

kubectl (kube control tool)

• Kubernetes CLI used to
  • Deploy and manage applications on a cluster
  • Get cluster related information
  • Get status of nodes
  • etc

kubectl run

Run an application on the cluster

kubectl run hello-minikube

kubectl cluster-info

Get information about the cluster

kubectl cluster-info

kubectl get nodes

List all nodes in the cluster

kubectl get nodes

Kubernetes Deployment Example

To run hundreds of applications across all nodes in the cluster, run a single command:

kubectl run my-web-app --image=my-web-app --replicas=1000