Journey to GIFEE - Philly ETE 2016

Slide 1

Slide 1 text

Brandon Philips @brandonphilips | brandon.philips@coreos.com The Journey to #GIFEE Intro to the Open Source

Slide 2

Slide 2 text

Secure the Internet MISSION

Slide 3

Slide 3 text

Separate Apps from OS STRATEGY

Slide 4

Slide 4 text

Make Servers Consistent STRATEGY

Slide 5

Slide 5 text

Tolerate Machine Failures STRATEGY

Slide 6

Slide 6 text

Make Servers Easy to Upgrade STRATEGY

Slide 7

Slide 7 text

Simplify Application Upgrades STRATEGY

Slide 8

Slide 8 text

No content

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

No content

Slide 11

Slide 11 text

No content

Slide 12

Slide 12 text

No content

Slide 13

Slide 13 text

No content

Slide 14

Slide 14 text

3 Application packaging Clustering Linux at scale

Slide 15

Slide 15 text

#GIFEE Borg/Omega Linux Chubby

Slide 16

Slide 16 text

#GIFEE Borg/Omega Linux Chubby

Slide 17

Slide 17 text

#GIFEE Borg/Omega Linux Chubby

Slide 18

Slide 18 text

Avoid single points of failure Design for constant updates Consistent environment Why build #GIFEE?

Slide 19

Slide 19 text

Why build #GIFEE? Design for constant updates

Slide 20

Slide 20 text

Application Packaging 1

Slide 21

Slide 21 text

Abstract away app from the OS OS App

Slide 22

Slide 22 text

No content

Slide 23

Slide 23 text

No content

Slide 24

Slide 24 text

Base software managed by CoreOS kernel systemd OpenSSH

Slide 25

Slide 25 text

Protect apps from each other Isolated network namespace Isolated file system namespace Mixed versions of dependencies eg. python 3.4 & python 2.7

Slide 26

Slide 26 text

$ sudo rkt run coreos.com/etcd:v2.0.0 $ sudo rkt run coreos.com/etcd:v2.0.0 \ --cpu=750m --memory=128M $ sudo rkt run --net=host coreos.com/etcd:v2.0.0 rkt run

Slide 27

Slide 27 text

Search container metadata Identify vulnerabilities Explain update actions Clair container security auditing

Slide 28

Slide 28 text

After scanning millions of containers we found that over 80% still had Heartbleed 80% Clair container security auditing

Slide 29

Slide 29 text

No content

Slide 30

Slide 30 text

In-Progress Universal Container Format Packaged Downloaded Verified

Slide 31

Slide 31 text

Linux at Scale 2

Slide 32

Slide 32 text

Patches to the OS and kernel are hard Retest after updates No automation SECURITY Dependency breakage Uptime risk APPLICATION

Slide 33

Slide 33 text

No content

Slide 34

Slide 34 text

Auto-updating browsers fixed security We got HTML5 at the same time

Slide 35

Slide 35 text

Atomic operating system updates

Slide 36

Slide 36 text

Atomic operating system updates

Slide 37

Slide 37 text

Clustering 3

Slide 38

Slide 38 text

Patches to the OS and kernel are hard Retest after updates No automation SECURITY Dependency breakage Uptime risk APPLICATION

Slide 39

Slide 39 text

Patches to the OS and kernel are hard No automation SECURITY Uptime risk APPLICATION

Slide 40

Slide 40 text

Operations Paradise Easy scale out Painless app upgrades Tolerant of machine failure

Slide 41

Slide 41 text

App Req/sec: 6,000 App Healthy: True

Slide 42

Slide 42 text

App Req/sec: 6,000 App Healthy: True

Slide 43

Slide 43 text

App Req/sec: 7,000 App Healthy: True

Slide 44

Slide 44 text

App Req/sec: 8,000 App Healthy: True

Slide 45

Slide 45 text

App Req/sec: 7,000 App Healthy: True

Slide 46

Slide 46 text

App Req/sec: 6,000 App Healthy: True

Slide 47

Slide 47 text

App Req/sec: 8,000 App Healthy: True

Slide 48

Slide 48 text

App Req/sec: 7,000 App Healthy: True

Slide 49

Slide 49 text

App Req/sec: 8,000 App Healthy: True

Slide 50

Slide 50 text

App Req/sec: 8,000 App Healthy: True

Slide 51

Slide 51 text

App Req/sec: 8,000 App Healthy: True

Slide 52

Slide 52 text

When do you need cluster coordination? Leader election Cluster-wide Semaphores Service discovery Dynamic configuration

Slide 53

Slide 53 text

Hard Computer Science Problem ?

Slide 54

Slide 54 text

Hard Computer Science Problem Chubby

Slide 55

Slide 55 text

A highly-available key value store for shared configuration and service discovery

Slide 56

Slide 56 text

No content

Slide 57

Slide 57 text

No content

Slide 58

Slide 58 text

No content

Slide 59

Slide 59 text

No content

Slide 60

Slide 60 text

No content

Slide 61

Slide 61 text

No content

Slide 62

Slide 62 text

No content

Slide 63

Slide 63 text

No existing “cloud native” solutions Simple HTTP + JSON APIs Dynamic reconfiguration Why build etcd?

Slide 64

Slide 64 text

Simple key/value “Distributed etc” Keys are versioned Changes can be watch

Slide 65

Slide 65 text

Cluster-wide reboot lock - locksmith Service discovery - vulcand, skydns Cluster orchestration - k8s, cloud foundry

Slide 66

Slide 66 text

Industry Adoption 500+ projects on Github

Slide 67

Slide 67 text

When do you need cluster coordination? Leader election Cluster-wide Semaphores Service discovery Dynamic configuration

Slide 68

Slide 68 text

Hard Computer Science Problem ?

Slide 69

Slide 69 text

Hard Computer Science Problem Chubby

Slide 70

Slide 70 text

A highly-available key value store for shared configuration and service discovery

Slide 71

Slide 71 text

Kubernetes is our choice for orchestration platform

Slide 72

Slide 72 text

App Req/sec: 8,000 App Healthy: True

Slide 73

Slide 73 text

App Req/sec: 8,000 App Healthy: True

Slide 74

Slide 74 text

App Req/sec: 8,000 App Healthy: True

Slide 75

Slide 75 text

App Req/sec: 8,000 App Healthy: True

Slide 76

Slide 76 text

Guides & Tools coreos.com/kubernetes kube-aws Cloud-configs

Slide 77

Slide 77 text

Scheduler gets work to the servers

Slide 78

Slide 78 text

$ scp app host:/opt $ ssh host systemd-run /opt/app

Slide 79

Slide 79 text

$ scp app host:/opt $ ssh host systemd-run /opt/app

Slide 80

Slide 80 text

$ fab deploy:app

Slide 81

Slide 81 text

$ fab deploy:app

Slide 82

Slide 82 text

$ fab deploy:app

Slide 83

Slide 83 text

$ fab deploy:gpu-app

Slide 84

Slide 84 text

$ fab deploy:gpu-app

Slide 85

Slide 85 text

$ fab deploy:gpu-app

Slide 86

Slide 86 text

Replication controllers drive the cluster to a desired state

Slide 87

Slide 87 text

pod env=prod app=web pod env=prod app=web pod env=prod app=web rc web-prod select(env=prod,app=web) count=1

Slide 88

Slide 88 text

pod env=prod app=web pod env=prod app=web pod env=prod app=web rc web-prod select(env=prod,app=web) count=1

Slide 89

Slide 89 text

pod env=prod app=web rc web-prod select(env=prod,app=web) count=1

Slide 90

Slide 90 text

pod env=prod app=web rc web-prod select(env=prod,app=web) count=5

Slide 91

Slide 91 text

pod env=prod app=web pod env=prod app=web pod env=prod app=web pod env=prod app=web pod env=prod app=web rc web-prod select(env=prod,app=web) count=5

Slide 92

Slide 92 text

Services find scheduled apps (pods) and load balance them

Slide 93

Slide 93 text

pod env=dev app=web pod env=test app=web pod env=prod app=web

Slide 94

Slide 94 text

pod env=dev app=web pod env=test app=web pod env=prod app=web service test.example.com select(env=dev,app=web) service beta.example.com select(env=test,app=web) OR select(env=prod,app=web) service example.com select(env=prod,app=web)

Slide 95

Slide 95 text

pod env=test app=web pod env=prod app=web pod env=prod app=web pod env=dev app=web pod env=test app=web pod env=prod app=web service test.example.com select(env=dev,app=web) service beta.example.com select(env=test,app=web) OR select(env=prod,app=web) service example.com select(env=prod,app=web)

Slide 96

Slide 96 text

pod app=foo,version=1 service foo.cluster.local select(app=foo)

Slide 97

Slide 97 text

pod app=foo,version=1 pod app=foo,version=2 service foo.cluster.local select(app=foo)

Slide 98

Slide 98 text

pod app=foo,version=1 pod app=foo,version=2 service foo.cluster.local select(app=foo)

Slide 99

Slide 99 text

Upstream rktnetes Auth/OIDC Node self-signed TLS

Slide 100

Slide 100 text

Scaling 15x scheduler performance 30k pods on 1k nodes SIG-scale

Slide 101

Slide 101 text

3 Application packaging Clustering Linux at scale

Slide 102

Slide 102 text

Sounds good, but... Is anyone successful with all this in prod?

Slide 103

Slide 103 text

Publically traded options exchange

Slide 104

Slide 104 text

Containers on CoreOS are powering ISE's high- throughput, low-latency financial exchange Running in production Bare metal & AWS Billions of transactions a day 150 million req/sec

Slide 105

Slide 105 text

No content

Slide 106

Slide 106 text

Secure the Internet MISSION

Slide 107

Slide 107 text

Separate Apps from OS STRATEGY

Slide 108

Slide 108 text

Make Servers Consistent STRATEGY

Slide 109

Slide 109 text

Tolerate Machine Failures STRATEGY

Slide 110

Slide 110 text

Make Servers Easy to Upgrade STRATEGY

Slide 111

Slide 111 text

Simplify Application Upgrades STRATEGY

Slide 112

Slide 112 text

No content

Slide 113

Slide 113 text

coreos.com/fest - @coreosfest May 9 & 10, 2016 - Berlin, Germany

Slide 114

Slide 114 text

No content

Slide 115

Slide 115 text

Build, Store and Distribute your Containers quay.io

Slide 116

Slide 116 text

Thank you! Brandon Philips @brandonphilips | brandon@coreos.com | coreos.com We’re hiring in all departments! Email: careers@coreos.com Positions: coreos.com/ careers