Very Large Development

How to run code review for
1000+ Open Source Developers

Joe Gordon

About Me

Very Large Development

As defined by ohloh by BLACK DUCK

“ Very large, active development team
Over the past twelve months, 1338 developers contributed new code to OpenStack. This is one of the largest open-source teams in the world, and is in the top 2% of all project teams on Ohloh.
For this measurement, Ohloh considers only recent changes to the code. Over the entire history of the project, 1808 developers have contributed.”

What is ?

Largest Open Source project?

(That I know of)

is 966,104 lines of python code
and OpenStack is 1,860,725!




Who is OpenStack

Deployment scale

The Data center

(Rackspace data center)

Development scale

By the Numbers


In January 2011

  • 61 Contributors
  • 71,181 lines of code

Pros and Cons
of using


  • fast to develop
  • approachable language
  • 'fun'


  • type checking
  • other static analysis
  • concurrency

Development Process

Life of a github Patch Today

  1. Fork repo
  2. Write code, test code and push to your github repo
  3. Submit a pull request
  4. travis-ci tests patch
  5. Patch is reviewed
  6. Patch is amended to address any negative comments
  7. Code is merged
  8. travis-ci runs on trunk

Where we started

Launchpad (bzr and all)

(3 years ago)

Review Process Evolution


Gate pep8 + unit tests

Gate pep8 + py26, py27, integration tests

Gate + Check


Life of an OpenStack Patch Today

  1. Code is written and tested locally
  2. Submitted for code review
  3. Code is automatically tested on submission
  4. Code is peer reviewed
  5. Patch is amended to address any negative peer reviews
  6. Code is approved (or rejected)
  7. Code is re-tested as it will be merged
  8. Code is merged



  • Never break trunk
    • Master branch is always green
    • Developers are never blocked on broken trunk
  • Transparency
  • Automate everything
  • Egalitarian
  • Be Strict. Reduce burden on reviewers


  • Work Flow
  • Testing
  • Integration Testing
  • Communication

Work Flow

  • Gerrit
  • git-review
  • Jenkins


  • tox
  • testrepository
  • Automated code quality checks: flake8+pep8+pyflakes+hacking


automated code quality checks

  • E1** indentation
  • E2** white space
  • E5** line length
  • F401 module imported but unused
  • F821 undefined name name
  • H103 correct apache 2 license
  • H30* import rules
  • H301 docstring should not start with a space

Integration Testing

  • devstack
  • zuul
  • recheck
  • elastic-recheck

Integration Testing


  • Measure integration tests in percent failure, not black and white
  • Hard to debug transient failures - 10s of MB of logs
  • Trivial to classify failures once have elastic search query


  • lauchpad
  • etherpad
  • pastebin
  • irc
  • mailing list

Suggestions for other tools?

Other ways we make development scale

Shrink Project Scope

OpenStack started with just two projects - Nova and Swift.

Now the same functionality is spread out over 6 projects.

Nova, Swift, Glance, Keystone, Cinder, Neutron

Thank You


Statistics from

Powered by reveal.js