10x deployment frequency using GitLab Flow and GitHub Actions on AWS ECS

English version of this slide
https://speakerdeck.com/tadashi0713/depuroipin-du-wo10bei-nisita-burantizhan-lue-togithub-actions-on-aws-ecs

#githubactions #devops #awsecs #cicd #gitlabflow

Transcript

1. 10x deployment frequency
   using GitLab Flow and
   GitHub Actions on AWS ECS
   Tadashi Nemoto
   

   View Slide

2. Introduction
   ● Tadashi Nemoto
   ● Exawizards, Inc.
   ● Platform Engineer (DevOps Engineer)
   ○ Introduce and improve CI / CD pipelines
   ○ Maintain production / develop environments
   ○ Introduce Test Automation
   

   View Slide

3. Have you introduced CI / CD ?
   

   View Slide

4. Are these CI / CD pipelines
   optimized for current
   product / organization?
   

   View Slide

5. State of DevOps Report 2019
   

   View Slide

6. State of DevOps Report 2018
   

   View Slide

7. State of DevOps Report 2018
   

   View Slide

8. Startup companies need speed to market
   

   View Slide

9. View Slide

10. Deployment Frequency(27 deploys / 2 weeks)
    10x Deployment Frequency
    

    View Slide

11. Outlines
    ● Previous CI /CD pipelines and deployment flows
    ● What I improved
    ○ Jenkins → GitHub Actions on AWS ECS
    ○ Git Flow → GitLab Flow
    ● Effect of improvements
    ● Challenges
    

    View Slide

12. Previous CI / CD pipelines
    and deployment flows
    

    View Slide

13. Previous CI / CD・Deployment flows
    ● Hashicorp Nomad on AWS
    ○ develop, staging, production environments
    ○ Difficult to create multiple development environments
    ● Git Flow as branching strategy
    ○ Usages are slightly different depending on teams
    ● Jenkins on AWS
    ○ Some deployments to production depend on us
    

    View Slide

14. Enable small, autonomous
    deployments
    

    View Slide

15. Increase deployment
    frequency
    

    View Slide

16. Improvement ①
    Jenkins → GitHub Actions on AWS ECS
    

    View Slide

17. Jenkins
    ● Maintenance cost
    ○ Update versions, plugins
    ○ Add machines, scale
    ○ Authority management, Security, etc.
    ● Need dedicated members or teams
    ● Not suitable for autonomous deployments
    

    View Slide

18. SaaS CI / CD tools
    

    View Slide

19. Limitation of deployment
    

    View Slide

20. GitHub Actions self-hosted runners
    

    View Slide

21. GitHub Actions self-hosted runners
    ● GitHub Actions provides self-hosted runners
    ● No additional charge
    ● Compatible both self-hosted and
    GitHub-hosted runners
    ○ Deploy → self-hosted runners
    ○ Test → GitHub-hosted runners
    ● No need to maintain the management part of a
    CI / CD workflow (like a Jenkins master instance)
    

    View Slide

22. GitHub Actions self-hosted runners on AWS ECS
    

    View Slide

23. https://techblog.exawizards.com/entry/2020/11/04/100000
    

    View Slide

24. Improvement ②
    Git Flow → GitLab Flow
    

    View Slide

25. Git Flow
    

    View Slide

26. Git Flow
    ● Works well for projects with fixed release timing
    ○ Mobile apps(1 release / 1~2 weeks)
    ● Less beneficial for projects API / Frontend projects
    ● More complex branch management, and not
    suitable for deploying to production frequently
    ○ Creating release branches, Hotfix branches, Tags
    

    View Slide

27. GitHub Flow
    

    View Slide

28. GitHub Flow
    Prod
    env
    ︖env
    ● Simpler branch management
    ○ master / feature branch
    ● Enable high deployment frequency
    ● Problems with
    verification environment
    ○ master branch = prod environment
    ○ staging environment︖
    ○ development environment︖
    ︖env
    

    View Slide

29. I want high deployment frequency
    while utilizing current environments
    (develop, staging, production)
    

    View Slide

30. GitLab Flow
    ● feature branch → merge to master branch
    ○ Same as GitHub Flow
    ● GitLab Flow allows to have branch
    for a release
    ○ master branch → staging env
    ○ production branch → production env
    ○ Merge to production branch
    for production release
    Staging
    env
    Prod
    env
    

    View Slide

31. Deploy to develop environment (workflow_dispatch)
    

    View Slide

32. Automatically generate release pull requests
    Staging
    env
    Prod
    env
    

    View Slide

33. https://techblog.exawizards.com/entry/2021/01/28/113812
    

    View Slide

34. Effect of improvements
    

    View Slide

35. Deployment Frequency(4 deploys / 1 month)
    

    View Slide

36. Deployment Frequency(27 deploys / 2 weeks)
    

    View Slide

37. Deployment Frequency(27 deploys / 2 weeks)
    10x Deployment Frequency
    

    View Slide

38. State of DevOps Report 2019
    

    View Slide

39. Small, autonomous deployments
    

    View Slide

40. Challenges
    

    View Slide

41. Continuous monitoring / improvement
    

    View Slide

42. PR based environments / drop self-hosted runners
    Staging
    env
    PR1 env
    PR2 env
    

    View Slide

43. Continuous Testing / Inspection
    

    View Slide

44. Conclusion
    

    View Slide

45. Message(Key takeaways)
    ● Important to improve deployment frequency,
    especially startups
    ● I achieved 10x deployment frequency
    by 2 improvements
    ○ Jenkins → GitHub Actions on AWS ECS
    ○ Git Flow → GitLab Flow
    ● Continuous monitoring / improvement
    

    View Slide