Azure ACS and Kubernetes

Alexander Zeitler

Microservices on Steriods

Alexander Zeitler

CEO / Solution Architect at PDMLab

Microsoft MVP / ASP.NET since 2004

ASP.NET Web API Advisor since 2011

Microsoft Azure Advisor since 2014

Organizer of Azure Meetup Karlsruhe

Using Docker since 2013

https://pdmlab.com

alexander.zeitler@pdmlab.com

@alexzeitler_

Agenda

  • Microservices
  • Challenges
  • Docker
  • Azure ACS / Kubernetes

Microservices

Microservices is a variant of the service-oriented architecture (SOA) architectural style that structures an application as a collection of loosely coupled services.

http://en.wikipedia.org/wiki/Microservices

Microservices-based architectures enable continuous delivery and deployment.

In a microservices architecture, services should be fine-grained and the protocols should be lightweight. The benefit of decomposing an application into different smaller services is that it improves modularity and makes the application easier to understand, develop and test.

The benefit of decomposing an application into different smaller services is that it improves modularity and makes the application easier to understand, develop and test. It also parallelizes development by enabling small autonomous teams to develop, deploy and scale their respective services independently

Example: E-Commerce

  • Web Front-End
  • Product Data Management (PDM)
  • Warehouse
  • Purchasing
  • Logistics
  • Payment
  • Billing
  • Identity

Principles

  • Focus on a single Domain
  • Two Pizza Rule
  • Autonomous
  • Small
  • Shared Nothing (Exclusive State)
  • Async

Advantages

  • Cross Platform
  • Easier Replacement / Refactoring
  • Decoupled Deployment
  • Code Ownership

Challenge: Complexity

Complexity

  • Significant Operations Overhead
  • Substantial DevOps Skills Required
  • Implicit Interfaces
  • Duplication Of Effort
  • Distributed System Complexity
  • Asynchronicity Is Difficult!
  • Testability Challenges

http://highscalability.com/blog/2014/4/8/microservices-not-a-free-lunch.html

Challenge: Deployment

Deployment

  • Fast
  • Reliable
  • Repeatable

Automate Deployment!

The perfect deployment?

Build, ship and run any application anywhere

http://upload.wikimedia.org/wikipedia/commons/c/c5/Britain_Delivers_the_Goods_in_Wartime-_Dock_Workers_in_Bristol,_England,_1940_D1224.jpga

http://i1.wp.com/www.chinasmack.com/wp-content/uploads/2013/09/historical-shanghai-photos-early-20th-century-14.jpg

vs...

https://homefaraway.files.wordpress.com/2012/04/april-24-2012-026.jpg

Put it in a container

Build

package application in a container image

Ship

Move container image to another machine

Run

Execute container(s) based on image

Any Application

Anywhere

Build

FROM ubuntu:16.04
MAINTAINER Alexander Zeitler <alexander.zeitler@pdmlab.com>
RUN apt-get update
RUN apt-get install -y nginx
RUN echo 'Hi, I am your father...erm in your container' \
>/usr/share/nginx/html/index.html
CMD [ "nginx", "-g", "daemon off;" ]
EXPOSE 80

Ship

Image name <username>/<reponame>
e.g. microsoft/dotnet


docker push
docker pull

https://hub.docker.com/

Run


docker run ubuntu echo hello world

Docker ecosystem

Engine

Compose

Hub

Machine

Swarm

Registry

Cloud

Containers in Azure

Azure Container Service

ACS Orchestrators

Docker Swarm

Mesos / DC/OS

Kubernetes

Kubernetes

Greek for "Helmsman"

The person who steers a ship

Invented by Google

Borg

Omega

Kubernetes == k8s

k8s Architecture

HTTPS

KV

Auth

Db

Log

Messaging

LB

Orders

HTTPS

KV

Auth

Db

Log

Messaging

LB

Orders

HTTPS

KV

Auth

Db

Log

Messaging

LB

Orders

Node 1

Node 2

Node 3

{ /*App*/}

Master

Master

Master

Node

Node

Node

Node

Node

Node

{ /*App*/}

Master

Master

Master

Node

Node

Node

Node

Node

Node

Deployment

Pods

Pods

Pods

Master

Node

Linux

Linux

Node

Linux

Node

Linux

Env

Env

Env

Env

k8s Masters

Master

Linux

Env

KV

apiserver

kube-apiserver

Front end to control plane

REST-API

JSON (manifest files)

 

KV Store

persistent

state config of the cluster

Uses etcd

CM

Controller Manager

Node Controller

Endpoints Controller

Namespace Controller

kube-scheduler

watches apiserver

Assigns work to nodes

 

Master

KV

apiserver

CM

kubectl

apiserver

Commands (JSON)

Node

Node

Node

k8s Nodes

Node

Linux

Env

Kubelet

kubernetes agent

registers node with cluster

watches apiserver

instantiates pods

reports to master

endpoint at port :10225

  /spec

  /healthz

  /pods

Container Runtime

pull images

start / stop containers

Docker or rkt

kube-proxy

Networking

Assign IP to Pod

All containers in a Pod share IP

LB across Pods in services

K

CR

P

Desired State

Master

KV

apiserver

CM

kubectl

apiserver

Manifest file (JSON/YAML)

Node

Node

Node

Master

KV

apiserver

CM

kubectl

apiserver

Manifest file (JSON/YAML)

Node

Node

Node

Pod

Pod

Pod

3 x nginx

nginx

nginx

nginx

Desired State:

Master

KV

apiserver

CM

kubectl

apiserver

Manifest file (JSON/YAML)

Node

Node

Node

Pod

Pod

Pod

3 x nginx

nginx

nginx

nginx

Desired State:

Master

KV

apiserver

CM

kubectl

apiserver

Manifest file (JSON/YAML)

Node

Node

Pod

Pod

3 x nginx

nginx

nginx

Pod

Desired State:

Pods

Unit of scale

Pod

Pods

Pod

link

net

trnsprt

IPC

mount

Environment

Namespaces

Network stack

...

n Containers

Scaling

Pod

Pod

Pod

Pod

Coupling

Pod

App

AI-Log

Pods are atomic

containers not up == pod n/a

Pod lifecycle

born - alive - dead

Pod Deployment

"manually"

Manifest

Services

Pod

Pod

Pod

Pod

Pod

10.244.2.4

10.244.2.5

10.244.2.6

10.244.2.6

10.244.2.6

Service

10.244.2.10

DNS: db.myapp

Labels

Pod

Pod

Pod

Pod

Pod

10.244.2.4

10.244.2.5

10.244.2.6

10.244.2.6

10.244.2.6

Service

10.244.2.10

DNS: db.myapp

p

p

p

p

p

p

db

db

lb

Deployments

Deployments

Manifest files (YAML/JSON)

 

Rolling updates + rollbacks

Multiple concurrent versions

   blue-green-deployment

   Canary releases

Demo

Thanks!

Azure ACS and Kubernetes

By Alexander Zeitler

Made with Slides.com

Azure ACS and Kubernetes

  • 933

More from Alexander Zeitler