Slide 1

Slide 1 text

Achieving repeatable, extensible and self serve infrastructure

Slide 2

Slide 2 text

2 tasdikrahman.me @tasdikrahman ● Product Engineer @ Gojek ● Contributor to oVirt ● Backpacker ● Weekend chef ● Chelsea FC!!

Slide 3

Slide 3 text

What does Gojek do? 3

Slide 4

Slide 4 text

4 Ref: gojek.io

Slide 5

Slide 5 text

What am I gonna talk about? 5

Slide 6

Slide 6 text

6 Ref: shutterstock.com

Slide 7

Slide 7 text

7 Ref: shutterstock.com Evolution of Infrastructure @ Gojek

Slide 8

Slide 8 text

Travelling back in time 8

Slide 9

Slide 9 text

Rapid Demand 9

Slide 10

Slide 10 text

How to deal with it? 10

Slide 11

Slide 11 text

Central Infrastructure Team 11

Slide 12

Slide 12 text

Intent? 12

Slide 13

Slide 13 text

Abstract out Infrastructure For Product Teams 13

Slide 14

Slide 14 text

Outcome? 14

Slide 15

Slide 15 text

Adhoc requests 15

Slide 16

Slide 16 text

“Measure what is measurable, and make measurable what is not so” - Galileo 16 Credits: biography.com

Slide 17

Slide 17 text

Service request tickets 17

Slide 18

Slide 18 text

18 Example service request in our ticket system by a team (names redacted)

Slide 19

Slide 19 text

19 Example service request to increase disk size (names redacted)

Slide 20

Slide 20 text

Number of service requests kept increasing with scale and more product groups coming in 20

Slide 21

Slide 21 text

21 Ref: gunshowcomic.com/648

Slide 22

Slide 22 text

How does one keep up with service requests? 22

Slide 23

Slide 23 text

Scale your team vertically and keep doing so 23

Slide 24

Slide 24 text

Sustainable? 24

Slide 25

Slide 25 text

Very hard to do, but mostly No 25

Slide 26

Slide 26 text

Eventually, we noticed we were becoming the bottleneck 26

Slide 27

Slide 27 text

Give access to someone from the product team? 27

Slide 28

Slide 28 text

Chances of Security loopholes 28

Slide 29

Slide 29 text

29 Ref: https://blog.codinghorror.com/the-broken-window-theory/

Slide 30

Slide 30 text

What do we do then? 30

Slide 31

Slide 31 text

Quick detour 31

Slide 32

Slide 32 text

Where did systems administration start? 32

Slide 33

Slide 33 text

Evolution of Automation at Gojek 33

Slide 34

Slide 34 text

Evolution of Automation at Gojek 34 ● Scripts ● Chef-cookbooks ● Rundeck ● Deployment scripts

Slide 35

Slide 35 text

Problems with the earlier solutions 35 ● Multiple ways around building and using automation ● Managing dependencies for the automation. Eg: people using gcloud/AWS

Slide 36

Slide 36 text

Problems with the earlier solutions 36 ● Lack of convention leading to meagre contributions to automation from devs. ● Adhoc way of managing access to tools like terraform, knife leading to stray accidents. ● No central platform for automation.

Slide 37

Slide 37 text

Number of tickets getting created still not decreasing 37

Slide 38

Slide 38 text

Clearing infrastructure debts 38

Slide 39

Slide 39 text

Moving from maintenance to innovation mode 39

Slide 40

Slide 40 text

Making infrastructure boring for product teams 40

Slide 41

Slide 41 text

Proctor: Our automation orchestrator 41 Ref: github.com/gojek/proctor

Slide 42

Slide 42 text

42

Slide 43

Slide 43 text

43

Slide 44

Slide 44 text

Installation 44

Slide 45

Slide 45 text

45 Helm all the way Reference value: stable/proctor-service/values.yaml

Slide 46

Slide 46 text

Automation using proctor 46

Slide 47

Slide 47 text

Sample proc to increase disk 47

Slide 48

Slide 48 text

Sample proc to increase disk 48

Slide 49

Slide 49 text

Scripts can be added by developers and they get added to proctor after our review 49

Slide 50

Slide 50 text

Sample procs in our ecosystem 50

Slide 51

Slide 51 text

Demo 51

Slide 52

Slide 52 text

Profit? 52

Slide 53

Slide 53 text

Outcome of having proctor? 53

Slide 54

Slide 54 text

Decrease in number of tickets which were mechanical in nature 54

Slide 55

Slide 55 text

Having terraform inside CI 55 +

Slide 56

Slide 56 text

But before that 56

Slide 57

Slide 57 text

Creating the gcloud project 57

Slide 58

Slide 58 text

58 Sample directory structure

Slide 59

Slide 59 text

59 .gitlab-yml for the gcloud project in gitlab

Slide 60

Slide 60 text

60

Slide 61

Slide 61 text

61 Plan and apply

Slide 62

Slide 62 text

Private terraform registry consisting of 90+ modules 62

Slide 63

Slide 63 text

Outcome? 63

Slide 64

Slide 64 text

Teams managing and provisioning their own infra with our best practices baked in terraform modules 64

Slide 65

Slide 65 text

OSS alternatives? 65

Slide 66

Slide 66 text

66 Reference: runatlantis.io/

Slide 67

Slide 67 text

Ideal state? 67

Slide 68

Slide 68 text

68 Ref: Google SRE book: Eliminating toil

Slide 69

Slide 69 text

Known caveats? 69

Slide 70

Slide 70 text

Deletion of infra 70

Slide 71

Slide 71 text

Teams forget what they are using 71

Slide 72

Slide 72 text

Lessons learnt? 72

Slide 73

Slide 73 text

Avoid premature automation 73

Slide 74

Slide 74 text

High service requests for product teams is a smell 74

Slide 75

Slide 75 text

No Big bang changes 75

Slide 76

Slide 76 text

Documentation should go hand in hand, would affect productivity directly 76

Slide 77

Slide 77 text

Reduce steps for onboarding to your tooling, lesser the better 77

Slide 78

Slide 78 text

Invisible infrastructure 78

Slide 79

Slide 79 text

Product managers in Infrastructure teams 79

Slide 80

Slide 80 text

Prioritizing on innovation 80

Slide 81

Slide 81 text

Links and References ● https://github.com/gojek/proctor ● https://blog.gojekengineering.com/olympus-terraforming-repeatabl e-and-extensible-infrastructure-at-go-jek-42ad5b0a4f9a ● https://learn.hashicorp.com/terraform/development/running-terrafor m-in-automation ● https://lethain.com/product-management-infra-engineering/ 81

Slide 82

Slide 82 text

82 @tasdikrahman tasdikrahman.me