Building systems with etcd @ SCALE x14

Slide 1

Slide 1 text

Fault Tolerant Infrastructure Building Systems with etcd @coreoslinux @brandonphilips

Slide 2

Slide 2 text

Brandon Philips CTO, CoreOS github.com/philips

Slide 3

Slide 3 text

What is CoreOS?

Slide 4

Slide 4 text

What is CoreOS?

Slide 5

Slide 5 text

No content

Slide 6

Slide 6 text

No content

Slide 7

Slide 7 text

What is CoreOS?

Slide 8

Slide 8 text

No content

Slide 9

Slide 9 text

The smartest way to run your container infrastructure. tectonic.com @tectonic

Slide 10

Slide 10 text

QUAY Secure hosting for private Docker repositories quay.io @quayio

Slide 11

Slide 11 text

Why build CoreOS?

Slide 12

Slide 12 text

you

Slide 13

Slide 13 text

you as a sw engineer

Slide 14

Slide 14 text

your with Ada.Text_IO; procedure Hello_World is use Ada.Text_IO; begin Put_Line("Hello, world!"); end; #include int main() { printf("Hello, world!\n"); } package main import "fmt" func main() { fmt.Println("Hello, world!") }

Slide 15

Slide 15 text

your container image

Slide 16

Slide 16 text

your /bin/java /opt/app.jar /lib/libc

Slide 17

Slide 17 text

your /bin/python /opt/app.py /lib/libc

Slide 18

Slide 18 text

your com.example.app d474e8c57737625c

Slide 19

Slide 19 text

your d474e8c57737625c Signed By: Alice

Slide 20

Slide 20 text

you as an ops engineer

Slide 21

Slide 21 text

your

Slide 22

Slide 22 text

your com.example.webapp x3

Slide 23

Slide 23 text

your com.example.webapp x3

Slide 24

Slide 24 text

your ??? com.example.webapp x3

Slide 25

Slide 25 text

How do we do it?

Slide 26

Slide 26 text

architecture in practice cluster operations

Slide 27

Slide 27 text

worker kubelet worker kubelet worker kubelet scheduler & API worker kubelet w ku t worker kubelet

Slide 28

Slide 28 text

machine configuration OS operations

Slide 29

Slide 29 text

distributed configuration cluster operations

Slide 30

Slide 30 text

github.com/philips/hacks/tree/master/etcd- demos

Slide 31

Slide 31 text

etcd

Slide 32

Slide 32 text

/etc distributed

Slide 33

Slide 33 text

open source software failure tolerant durable watchable exposed via HTTP runtime reconfigurable

Slide 34

Slide 34 text

Data Store API -X GET Get Wait -X PUT Put Create CAS -X DELETE Delete CAD

Slide 35

Slide 35 text

clusters etcd basics

Slide 36

Slide 36 text

Typical Cluster Leader Follower

Slide 37

Slide 37 text

API etcd basics

Slide 38

Slide 38 text

fault tolerance etcd basics

Slide 39

Slide 39 text

Available Leader Follower

Slide 40

Slide 40 text

Available Leader Follower

Slide 41

Slide 41 text

Available Leader Follower

Slide 42

Slide 42 text

Unavailable Leader Follower

Slide 43

Slide 43 text

leader fault tolerance etcd basics

Slide 44

Slide 44 text

Available Leader Follower

Slide 45

Slide 45 text

Available Leader Follower

Slide 46

Slide 46 text

Temporarily Unavailable Leader Follower

Slide 47

Slide 47 text

Available Leader Follower

Slide 48

Slide 48 text

Unavailable Leader Follower

Slide 49

Slide 49 text

wal, snapshots, backups etcd durability

Slide 50

Slide 50 text

discovery, static etcd bootstrap

Slide 51

Slide 51 text

$ curl discovery.etcd.io/new?size=5 discovery.etcd.io/6eadeac2

Slide 52

Slide 52 text

discovery

Slide 53

Slide 53 text

discovery

Slide 54

Slide 54 text

discovery

Slide 55

Slide 55 text

Leader Follower discovery

Slide 56

Slide 56 text

live addition and removal etcd reconfig

Slide 57

Slide 57 text

Leader Follower

Slide 58

Slide 58 text

Leader Follower

Slide 59

Slide 59 text

Leader Follower

Slide 60

Slide 60 text

etcd apps

Slide 61

Slide 61 text

reboot locksmith etcd apps

Slide 62

Slide 62 text

No content

Slide 63

Slide 63 text

No content

Slide 64

Slide 64 text

Cluster Wide Reboot Lock ● Need to reboot? Decrement the semaphore key atomically with etcd. ● manager.Reboot() and wait... ● After reboot increment the semaphore key in etcd atomically.

Slide 65

Slide 65 text

skydns etcd apps

Slide 66

Slide 66 text

vulcand etcd apps

Slide 67

Slide 67 text

No content

Slide 68

Slide 68 text

confd etcd apps

Slide 69

Slide 69 text

pulling it together kubernetes

Slide 70

Slide 70 text

k8s/mesos/etc scheduler scheduling

Slide 71

Slide 71 text

getting work to servers scheduling

Slide 72

Slide 72 text

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

Slide 73

Slide 73 text

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

Slide 74

Slide 74 text

$ fab deploy:app

Slide 75

Slide 75 text

$ fab deploy:app

Slide 76

Slide 76 text

$ fab deploy:app

Slide 77

Slide 77 text

$ fab deploy:collector-app

Slide 78

Slide 78 text

$ fab deploy:collector-app

Slide 79

Slide 79 text

$ fab deploy:collector-app

Slide 80

Slide 80 text

$ fab deploy deploy:collector-app

Slide 81

Slide 81 text

$ fab lowest-loadaverage

Slide 82

Slide 82 text

$ fab lowest-loadaverage host1

Slide 83

Slide 83 text

$ fab lowest-loadaverage host1 $ fab -H host1 deploy:job

Slide 84

Slide 84 text

You Scheduler API Scheduler Machine(s)

Slide 85

Slide 85 text

while true { todo = diff(desState, curState) schedule(todo) }

Slide 86

Slide 86 text

while true { todo = diff(desState, curState) schedule(todo) }

Slide 87

Slide 87 text

while true { todo = diff(desState, curState) schedule(todo) }

Slide 88

Slide 88 text

while true { todo = diff(desState, curState) schedule(todo) }

Slide 89

Slide 89 text

dns, LBs, k8s labels services

Slide 90

Slide 90 text

flexible service discovery k8s labels

Slide 91

Slide 91 text

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

Slide 92

Slide 92 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)