Get back to sleep with Infrastructure as Code

Transcript

1. Get back to sleep with Infrastructure as Code
2. Once upon a time...
3. Two innocent kids decided to organize a workshop on Docker Swarm...
4. ...and they had to create tens of VMs by hand for the attendees to use.
5. It took them a full-time day, ...
6. ...but it was totally worth it.
7. It was the necessary pain, for the enormous gain; ...
8. a stellar workshop!
9. And the day of the 6-hour workshop came.
10. They were super excited.
11. But at the middle of the workshop...
12. ...they realised they had misconfigured their VMs.
13. Nothing would work!
14. And that was it...
15. ...they didn’t have enough time to fix the misconfigured VMs.
16. The workshop sunk.
17. Attendees left — rightfully — furious.
18. The company refunded all tickets.
19. The reputation of the company got obliterated.
20. Finally, it shut operations down in just a month.
21. That’s sad.
22. Also that never happened.
23. Because we did nothing by hand.
24. We used Terraform.
25. It’s good. You should use it too.
26. Thanks for your time!
27. loljk
28. </bad-joke>
29. Hi, I am Paris Kasidiaris ...and I desperately beg you to follow me on Twitter (@pariskasid)
30. Paris Kasidiaris I co-founded SourceLair (sourcelair.com). I co-organize the Docker Athens User Group (docker.gr). I HATE TOUCHING OUR PRODUCTION INFRASTRUCTURE.
31. Let’s talk about infrastructure
32. What is infrastructure?
33. It’s all resources needed for running our applications.
34. Computers From VMs in the cloud to bare metal
35. Network interfaces From floating IPs, SDN interfaces etc.
36. Storage Block storage, object storage etc.
37. Let’s talk about code
38. What is code?
39. Machine readable text.
40. What is Infrastructure as Code?
41. Resources described as machine readable text.
43. Infrastructure as Code building blocks
44. Declarations
46. State
48. Operator
49. The operator is the program that will give flesh and bones to machine readable text
50. Example
52. $ terraform plan Refreshing Terraform state in-memory prior to plan... ------------------------------------------------------------------------ An execution plan has been generated and is shown below. Resource actions are indicated with the following symbols: + create Terraform will perform the following actions: + digitalocean_droplet.workshop_node_vm[0] id: <computed> backups: "false" disk: <computed> image: "ubuntu-18-04-x64" [Truncated] Plan: 3 to add, 0 to change, 0 to destroy.
53. $ terraform apply digitalocean_droplet.workshop_node_vm.0: Still creating... (10s elapsed) digitalocean_droplet.workshop_node_vm.2: Still creating... (10s elapsed) digitalocean_droplet.workshop_node_vm.1: Still creating... (10s elapsed) digitalocean_droplet.workshop_node_vm.0: Still creating... (20s elapsed) digitalocean_droplet.workshop_node_vm.1: Still creating... (20s elapsed) digitalocean_droplet.workshop_node_vm.2: Still creating... (20s elapsed) digitalocean_droplet.workshop_node_vm.2: Still creating... (30s elapsed) digitalocean_droplet.workshop_node_vm.0: Still creating... (30s elapsed) digitalocean_droplet.workshop_node_vm.1: Still creating... (30s elapsed) digitalocean_droplet.workshop_node_vm[1]: Complete after 36s (ID: 1380764) digitalocean_droplet.workshop_node_vm[0]: Complete after 36s (ID: 1380765) digitalocean_droplet.workshop_node_vm[2]: Complete after 37s (ID: 1380770) Apply complete! Resources: 3 added, 0 changed, 0 destroyed.
54. $ terraform destroy An execution plan has been generated and is shown below. Resource actions are indicated with the following symbols: - destroy Terraform will perform the following actions: - digitalocean_droplet.workshop_node_vm[0] - digitalocean_droplet.workshop_node_vm[1] - digitalocean_droplet.workshop_node_vm[2] Plan: 0 to add, 0 to change, 3 to destroy. [Truncated] Destroy complete! Resources: 3 destroyed.
55. There are plenty of tools to choose
56. AWS CloudFormation Azure ARM templates Openstack Heat Ansible
57. Why Infrastructure as Code?
58. Reason #1: Control risk
59. Oh shit. I think I just nuked a production server by mistake. | Paris Kasidiaris, circa 2017
60. We cannot afford mistakes in production infrastructure.
61. People were, are and will always be doing mistakes though.
62. Instead of trying to change human nature, we should use systems to save us from ourselves.
63. Reason #2: Save time
64. For automation to make sense, we need volume.
65. Updating 4 VMs by hand vs Terraform doesn’t say much.
66. Updating 40 VMs though...
67. ...or even 4 VMs, but 10 times in a day
68. Now we are talking...
69. Reason #3: Auditing
70. Do you remember what is the last tuning you performed on your app servers?
71. Or when or why you did it at all?
72. I guess not.
73. Me neither (I am not a freak).
74. Can you look for it though?
75. We can.
76. Still looking for a RealWorld™ use case?
77. True Story™ ahead
78. Last month we migrated all of SourceLair’s infrastructure
79. Our infrastructure 1. 5 VM types (Load Balancer, App VMs, Database VMs, Swarm Managers) 2. Multiple block storage devices 3. Multiple terabytes of customer data 4. 30 Docker Swarm Services
80. One simply does not migrate such a payload by crossing their fingers.
81. We had to iterate on provisioning, testing and destroying infra.
82. This is not fast, or cheap.
83. If we did all of these by hand, we would still be testing things.
84. And we would still be paying more than ⨉2 for infrastructure.
85. Instead, we used Terraform.
86. We provisioned, tested and destroyed multiple clusters in minutes.
87. This gave us the confidence to perform such a migration in a fraction of time.
88. So, the day we migrated all of our infrastructure...
89. We went back to sleep.
90. And slept tight — really, really tight!
91. One more thing...
92. How does one convert a pseudocode file to Python?
93. Just add the “.py” extension to the file name.
94. </worse-joke>
95. Learn more about how we mess with infrastructure!
96. https://2hog.codes
97. Thank you! Follow me on Twitter: @pariskasid
98. Questions?