Continuous Delivery Why / How / When ?

My speech on AgileBaseCamp 2012 "From Idea to Product"

Transcript

1. Why / How / When
   Do I need
   CONTINUOS
   DELIVERY?
   

   View Slide

2. Product Developer, E-conomic (Ciklum)
   Trainer, XP Injection
   beletsky.net
   @alexbeletsky
   github.com/alexanderbeletsky
   

   View Slide

3. Why
   How
   When
   

   View Slide

4. The story of one project
    Very simple GTD web application
    .NET technological stack based
    VPS hosted
    Nothing fancy, seriously
   

   View Slide

5. •
   Prepare release branch and merge all required changes there
   •
   Update version in uppercut.config
   •
   Commit changes to SCM
   •
   Run build.bat
   •
   FTP package to deployment server
   •
   RDP to deployment server
   •
   Unpackage .zip content to temp folder
   •
   Manually backup staging database
   •
   Stop Stage Web site in IIS manager
   •
   Run migration scripts for staging database
   •
   Run deployment scripts for staging environment
   •
   Run Stage Web site in IIS manager
   •
   Test manually that on staging server, that build works fine
   •
   If something missed (note it is 60% of all cases) go to 1
   •
   Manually backup production database
   •
   Stop Production Web site in IIS manager
   •
   Run migration scripts for production database
   •
   Run deployment scripts for production environment
   •
   Run Production Web site in IIS manager
   •
   Test manually that on production server, that build works fine
   Delivery Script ...
   

   View Slide

6. If it FAILS in a MIDDLE
   — REPEAT
   Lucky day: 0.5 h
   Bad day: 2 h
   

   View Slide

7. Read those figures as:
   My customers suffer for 2 hours once in
   2 weeks, just becase I do update
   

   View Slide

8. And by the way...
    It's just takes too much developers time
    It's boring
    Sooner or later you start to hate
   deployments
   

   View Slide

9. For bussiness people and
   managers
    “Time-to-Market” factor is very low
    Feedback cycle is very long
    Customer dissatisfaction is very high
   

   View Slide

10. Why
    How
    When
    

    View Slide

11. Let's define our
    TERMINOLOGY
    

    View Slide

12. Deployment != Delivery
    Deployment: product IS availble
    Delivery: product IS availble for users
    

    View Slide

13. Deployment first
    Delivery after
    

    View Slide

14. All you have to know,
    the secret formula of
    Continuous Deployment
    

    View Slide

15. (Integration + Deployment) *
    Continuous = Continuous
    Deployment
    

    View Slide

16. Integration is for:
    fetching latest sources
    build all product configuration items
    run unit tests
    run functional tests
    generate documentation
    package all artifacts
    Deployment is for:
    update test servers
    

    View Slide

17. 1. Build and deploy locally with one click
    2. Define SCM model based
    3. Run it continuosly
    Recipe of Success:
    

    View Slide

18.  Build and tests execution
     Binaries versioning
     SCM labeling
     Database migration
     Web application deployment
    Step 1 - Build and Deploy
    

    View Slide

19.  Build and tests execution
     Binaries versioning
     SCM labeling
     Database migration
     Web application deployment
    Step 1 - Build and Deploy
    Solved by Chuck Norris tools
    

    View Slide

20. UppercuT batch build framework
    RoundhousE database migrations
    DropicK application deployment
    Meet Chuck Norris Tools
    

    View Slide

21. UppercuT
    Easy to use
    Configurable by XML
    Supports config for different environment
    RoundhousE
    Easy to use
    Supports MS SQL, MySQL, Postgress
    Migration by SQL Scripts
    DropkicK
    Deployment script as C# code
    Support for services and sites
    Different plans
    

    View Slide

22. Available on a Github
    

    View Slide

23. > deploy.bat LOCAL
    > deploy.bat STAGING
    > deploy.bat PRODUCTION
    Step 1 - Build and Deploy
    ACCOMPLISHED
    

    View Slide

24.  DVCS are simply rule (Git, HG)
     TRUNK is production ready
     Keep interations in branches
     Keep features in branches
    Step 2 - Define SCM model
    

    View Slide

25. Ideal branches count = 2
    master
    develop
    Typical branches count >= 2
    master
    develop
    release
    hotfix
    

    View Slide

26. Branch per Environment Layout
    

    View Slide

27. Step 2 - Define SCM model
    ACCOMPLISHED
    > git checkout develop
    > git merge --no-ff myfeature
    > git branch -d myfeature
    > git push origin develop
    

    View Slide

28.  Availability of Build Server
     SCM build triggering
     Automatically run deployment script
     Roll out application to production
    Step 3 - Run it continuously
    

    View Slide

29.  Fork of famous Hudson project
     Open source
     Java based
     Easy start, easy go
     Tons of available plugins
    Say 'Hello' to Jenkins
    

    View Slide

30. All configuration could be
    done in UI
    

    View Slide

31. Jenkins Instance is deployed
    for each environment
    

    View Slide

32. Build/Deployment continuosly
    triggered by SCM
    

    View Slide

33. As soon as Staging “Looks
    Alright”, deploy to production
    

    View Slide

34. > git checkout master
    > git merge develop
    > git tag -a 1.2
    > git push origin master
    Deployment is nothing more
    as pushing changes to
    origin/master
    

    View Slide

35. Step 3 - Run it continuosly
    ACCOMPLISHED
    Changes are picked up, built, tested
    and deployed automatically
    

    View Slide

36. Why
    How
    When
    

    View Slide

37. As sooner as better
    

    View Slide

38. And?
    

    View Slide

39. Results:
     Going live time improved 45x
     Site down time reduced 300x
     No more iterations, Kanban
     Staging is updating with every push
     Setup and forget
    

    View Slide

40. THANK YOU!
    beletsky.net
    @alexbeletsky
    github.com/alexanderbeletsky
    

    View Slide