Continous Delivery:

Reliable software development through automation

Jesse Szwedko

j.szwedko@modcloth.com

@jesse_szwedko

Location: University of Pittsburgh

Date: 4/4/2012

Audience: Those with no or little background in continuous delivery (version control, automated testing, continuous integration, etc).

This talk: github.com/jszwedko/continuous-delivery-pitt-tasks (slides are on gh-pages branch)

modcloth

Presenter Notes

Agile

agile

Continuous feedback

Presenter Notes

  • Waterfall is "similar" just with a longer timeline
  • Flikr deploys 10 times a day

Continuous Delivery

Continuous Delivery (CD) is a design practice used in software development to automate and improve the process of software delivery. - Wikipedia

Presenter Notes

  • Anytime you are doing something manual, automate it
  • Rule of 3s

Automation

Presenter Notes

continuous delivery diagram

Presenter Notes

Outline

Process Evolution

  • Cowboy coding
  • Version control
  • Local development
  • Automated deploys
  • Automated testing
  • Continuous integration

Complete example

Presenter Notes

Tasks Application

tasks application

  • Allow the user to create ToDos
    • Create
    • Read
    • Update
    • Delete
    • List

Presenter Notes

  • Simple CRUD application
  • Focus on process
  • Show running live

ruby

Rails

Presenter Notes

  • Focus is on process, not technology
  • Please ask any technology questions

Feature request: As a user, I should be able to add a description for each todo

Presenter Notes

Cowboy Coding

Presenter Notes

Do it live

Presenter Notes

Do it live example

Presenter Notes

  • Add feature on remote box
  • Make syntax error
  • Revert to working version manually
  • Show broken site (motivate version control and developing locally)
  • Branch: initial-implementation

Disadvantages

  • Broken code is immediately visible to end user
  • Can not debug code (reveal application secrets)
  • End user sees a lot of volatility
  • Cannot easily undo changes

Presenter Notes

Outline

Process Evolution

  • Cowboy coding
  • Version control
  • Local development
  • Automated deploys
  • Automated testing
  • Continuous integration

Presenter Notes

Version control

Track changes to source code

Presenter Notes

  • Started with making copies of files
  • Unable to get back to stable state after changes

github

Presenter Notes

  • See source code at any point in time
  • Source code example (show commit, checkout of older version)
  • Image from http://programmers.stackexchange.com/questions/35074/im-a-subversion-geek-why-should-i-consider-or-not-consider-mercurial-or-git-or
  • Demonstrate commit and push to github

Version control example

Presenter Notes

  • Demonstrate commit, log, revert, checkout
  • Demonstrate needing to see current changes

Advantages

  • Can deploy code at any point in history (e.g. to rollback to stable version)
  • Can see why changes were performed
  • Allow for collaboration

Presenter Notes

  • Imagine collaborating on code

Still, we need to develop in a way where we can experiment

Presenter Notes

Outline

Process Evolution

  • Cowboy coding
  • Version control
  • Local development
  • Automated deploys
  • Automated testing
  • Continuous integration

Presenter Notes

Develop locally

  • Develop features, fix bugs, experiment locally
  • Push changes to server only after you are done
  • Consider using a virtual machine if your development enviroment is too different from the production environment

Presenter Notes

  • Use vagrant to spin up machine similar to "production" (especially when on Windows)
  • Will not quite be push button, but we will get there

Feature request: As a user, I should be able to add a description for each todo

Presenter Notes

Developing locally example

Presenter Notes

  • Develop adding of description text field
  • Push to github (show on Github)
  • git pull on server
  • Broken because of data migration!
  • Are not updating JSON builders

Advantages

  • Can develop feature in increments, deploy when finished
  • Allows for debugging
  • Experimentation

Presenter Notes

continuous delivery diagram

Presenter Notes

Closer, but multi-step deployments introduce risk

Presenter Notes

Deployment tasks

  • Copy code
  • Migrate database
  • Restart web server

Presenter Notes

Outline

Process Evolution

  • Cowboy coding
  • Version control
  • Local development
  • Automated deploys
  • Automated testing
  • Continuous integration

Presenter Notes

Automated deployments

  • Specify each step of the deploy as code
  • Deterministic
  • Idempotent
  • Should be able to rollback

deploy

Presenter Notes

  • Rule of 3s

capistrano

Presenter Notes

  • Automated deployment tool (primarily used for Ruby applications)
  • Other examples: ansible, chef, bash

Automated deployment example

Presenter Notes

  • Review capistrano configuration
  • Delete app/current
  • Deploy
  • Motivate need for tests

Advantages

  • Eliminate risk of multi-step deployments
  • Deploying to multiple servers
  • Automation as documentation

Presenter Notes

continuous delivery diagram

Presenter Notes

How do we make sure the application is stable before each deploy?

Presenter Notes

  • Poke the application manually (does not scale)
  • Stop regressions

Outline

Process Evolution

  • Cowboy coding
  • Version control
  • Local development
  • Automated deploys
  • Automated testing
  • Continuous integration

Presenter Notes

Automated Tests

  • Describe what the code should do
  • Catch regressions

Presenter Notes

MiniTest

Presenter Notes

  • Automated testing framework
  • Alternatives include RSpec, TestUnit

Feature request: As a user, I should be able to add a description for each todo

Presenter Notes

Automated testing example

Presenter Notes

  • Add description, but break title
  • Add view spec for show template to assert fields
  • Break test
  • Fix test

But what if we forget to run the tests?

Presenter Notes

Outline

Process Evolution

  • Cowboy coding
  • Version control
  • Local development
  • Automated deploys
  • Automated testing
  • Continuous integration

Presenter Notes

Travis CI

Presenter Notes

Continuous integration server

"Build" is triggered by commit or push to VCS

  • Run set-up tasks
  • Run tests
  • Build artifacts

Presenter Notes

  • git log w/ branches ("alternative timelines")
  • Github model uses pull requests
  • Branch
  • Add .travis.yml

Feature request: As a user, I should be able to add a description for each todo

Presenter Notes

CI Example (full example)

Presenter Notes

  • Create branch
  • Add description
  • Run tests
  • Push to github
  • Wait for Travis
  • Merge
  • Deploy

Advantages

  • Running tests happens automatically (you can trust other pull requests)
  • Running application under many different environments
  • Kick off longer test runs in the background

Presenter Notes

Conclusion

Presenter Notes

continuous delivery diagram

Presenter Notes

Mistakes we knew we were making

  • Skipped infrastructure configuration
  • You should try to avoid "building" on the server
    • E.g. we did asset compilation (translating of "higher" level languages to CSS and JS)
  • You should have a test stage target for deployment that is identical to production
    • Some issues are environment specific

Presenter Notes

Questions

Jesse Szwedko

j.szwedko@modcloth.com

@jesse_szwedko

This talk: github.com/jszwedko/continuous-delivery-pitt-tasks (slides are on gh-pages branch)

Further reading: Continuous Delivery: Reliable Software Releases through Build, Test, and Deployment Automation

modcloth

Presenter Notes

Technologies

Used here:

Others:

Presenter Notes

Devops Days

devopsdays

Presenter Notes