Continuously delivering continuous delivery

The rise and acceptance of Platform as a Service (PaaS) is a key tool in accelerating the journey to continuous delivery of your apps into production with the regularity your business requires. However, all too often the deployment and maintenance of the platforms themselves are dogged with the practices and delays that the CD of apps are all too familiar.

This session will serve as both an introduction and overview of the power of PaaS platforms, in this case Pivotal Cloud Foundry, with a selection of both application and platform case studies, continuously deployed all the way to production, in business critical and highly regulated environments.

Whether you're dealing with apps or globally distributed platforms, we hope to demonstrate that small batch sizes and continuous deployment are not only achievable but essential; no more 2am releases, no more working the weekends during "maintenance windows"

Speaker Bio: Ben Dodd has spent the last 6 years continuously deploying swarms of Microservcies to Platform as a Service environments. It’s been a journey fraught with challenges, mistakes and dead ends. Experience with organisations of all sizes has given him valuable insight into the value of true DevOps and Continuous Deployment in delivering and demonstrating confidence in hugely complex platforms, distributed globally and trusted with an organization's Crown Jewels. He is a regular speaker on topics such as Concourse CI, Cloud Foundry, continuous delivery and cloud native transformation within the enterprise

Transcript

1. Continuously delivering continuous delivery
2. Hypothesis: The only way to effectively manage high risk platforms is using continuous deployment
3. MY PaaS Basics
4. What is your target unit of currency?
5. An app.
6. What is the relationship you want to have with your environments?
8. Casual.
9. Where do you want to spend your time?
10. Adding app features that users want.
11. Terminology
12. So - what is a: BOSH?
13. So - what is a: Concourse CI?
14. So - what is a: Cloud Foundry?
15. I don’t care how, just run my app $ cf push myapp
16. $ cf scale -i 100 myapp SH!T, it’s a hit - I need more horsepower
17. $ cf create-service... I need a DB and but don’t want to manage it
18. $ cf create-space... I want orgs and spaces so I can safely share a platform
19. I want to build Pipelines now! … jobs: - name: job-deploy-myapp plan: - {get: resource-myapp, trigger: true} - put: resource-deploy-myapp params: manifest: myapp/manifest.yml path: myapp
20. Awesome! Let’s get us some PaaS
21. Problem 1: Platform work is manual work
23. Problem 2: We are a super special snowflake!
24. Problem 3: We are a [risky business]!
25. Problem 3: We are a [risky business]! PaaS == Platform as a Silo
26. Problem 4: We need to scale & update Updates every 4 days; 2 weeks to test each… So, when do I get my new feature?
27. Problem 5: It's expensive!
28. Deploying a PaaS Observations in the field 1. Organisations create their own PaaS 2. Organisations use an existing PaaS but deploy it manually 3. Organisations test a platform once when they go live 4. Organisations have a massive outage & can’t recover in time 5. Organisations need to update multiple times a week 6. Organisations need to scale by X; making 1-5 X times worse!
29. Examples
30. Apps.
31. Global, open-source & multi-cloud
32. Risk == Long feedback loops
35. Not an app.
36. Global bank Top 10 dev. houses
37. Zooportraits.com == awesome!
38. Examples Super Mega Awesome
39. Examples Super Mega Awesome Per pair environments
40. Examples Super Mega Awesome Destroyed daily
41. Examples Super Mega Awesome Pre-production continuously deployed
42. Examples Super Mega Awesome Next, production
43. Examples Super Mega Awesome Constant updates from the vendor
44. Examples Super Mega Awesome All config as code
45. Examples Super Mega Awesome Rolling deployment for platform & services
46. Examples Super Mega Awesome Test driven compliance
47. Examples Super Mega Awesome Test driven resilience
49. Continuously Delivering CD Finally, some top tips! 1. We kept a high line & stuck to our guns 2. We did TDD [using BATS] 3. We focused on the What? & Why?, not the How? it worked before 4. We didn’t have “prod” & “staging” named envs 5. We deployed in sociable times 6. We constantly optimised for short feedback loops
50. Thanks @bendodd @armakuniHQ ben.dodd@armakuni.com