Copyright (c) Shopkick 2017

See LICENSE for details.

Sky

Sky enables developers to build modern Python applications and microservices, with built-in support for Anaconda dependencies, Docker packaging, and DCOS deployment.

Quick Start

Download the newest version for your platform here: Sky releases.

chmod a+x ~/Downloads/sky-[ver]
alias sky=~/Downloads/sky-[ver]
sky --help

Installing pre-release/dev versions

git clone USERNAME@GITLAB_URL:REPO
cd sky
virtualenv build
build/bin/pip install -e .
alias sky="/path/to/sky/build/bin/python -m opensky"
export SKY_SITE_CONFIG=http://GITLAB_URL/REPO
sky pypier install sk-tunnel

Working on sky-based components

Next, clone the git repo of the component you'd like to develop on, and initialize the environment:

$ git clone git@GITLAB_URL:REPO
$ cd urlshortener
$ sky setup

This will download and set up any necessary library and service dependencies, from lib to mysql. Once it's done (5-15 minutes), you're ready to go!

$ sky start

This starts the component, and any backend service it needs. Future runs of sky setup are only necessary if the sky.yaml configuration is changed.

For more commands: sky --help!

                                                                                  .
       _____ _             _____                                          _     \ _ /
      / ____| |           / ____|                                        | |  -= (_) =-
     | (___ | | ___   _  | |     ___  _ __ ___  _ __ ___   __ _ _ __   __| |    /   \
      \___ \| |/ / | | | | |    / _ \| '_ ` _ \| '_ ` _ \ / _` | '_ \ / _` |    __'  _
      ____) |   <| |_| | | |___| (_) | | | | | | | | | | | (_| | | | | (_| |  _(  )_( )_
     |_____/|_|\_\\__, |  \_____\___/|_| |_| |_|_| |_| |_|\__,_|_| |_|\__,_| (_   _    _)
   __   _          __/ |                                                       (_) (__)
 _(  )_( )_       |___/             __   _           ____       _
(_   _    _)                      _(  )_( )_       |__\_\_o,___/ \
  (_) (__)                       (_   _    _)      ([___\_\_____-\'---</shopkick/<
                                   (_) (__)        | o'

FAQ

How do I split out a coreservice from the main shopkick repo?

See the migration guide next to this document.

How do I make changes in applications and libraries at the same time?

Codeveloping applications and libraries is a common and necessary practice here at shopkick. sky has special support for modifying applications (e.g., urlshortener) and library dependencies at the same time (e.g., lib). Always better to validate locally before pushing!

This is the procedure for codevelopment in sky:

1. Clone the lib repo (e.g. to your home directory)
2. In your application repo, cp sky.yaml sky.local.yaml
3. In sky.local.yaml, library_deps.sky.[library name], change the value from the git remote to the path to your local clone (e.g., "~/[library_name]" if you cloned to your home directory)
4. Restart the application. Note that rerunning sky setup is not necessary for this change to take effect.

Background

Sky was created to address the unique needs of shopkick's development process, present and future. Specifically we wanted to:

• Speed up development-deployment cycles through independent component deploys
• Reduce variance between development and production environments
• Upgrade to modern operating system, Python, and Thrift versions
• Enable a path forward for incremental updates (and usage) of libraries
• Emphasize reproducible development, minimize setup time, and eliminate "works on my machine"
• Leverage new open-source development tools (Docker, DCOS, GitLab)

The solution

Local development continues on Macs, a necessary evil for iOS app development. However, server components are run and tested inside a CentOS 7 + Python 2.7 environment using Mac Docker support. Production images are built and tested using GitLab CI from the same Dockerfile used locally, ensuring production and local work the same.

sky wraps Docker, conda, pip, DCOS, GitLab, and even certain git tasks so that developers can focus on development. sky becomes a unified entrypoint, replacing manage_backends.py, various test runners, make, scons, and more. sky turns new engineer onboarding from a 1-3 day semi-manual process into a 1-3 hour process, fully automated.

The heaviest dependencies shift from being built from source to being installed from binary artifacts. We leverage RPMs, conda packages, pip wheels, and Docker images to reduce build times to the absolute minimum. These artifacts are explicitly versioned and pinned to eliminate divergences between local build environments.

Individual GitLab repositories offer more CI and documentation opportunities. DCOS enables analogous gains for operations, some of which is exposed through sky. Developers trigger deployments using GitLab CI, which in turn uses sky to do deployments on DCOS. Developers can check on their DCOS configuration and status using sky, as well.

sky development and release also takes place on GitLab. It is shipped as a single-file executable, and supports plugins for easy extensibility. sky itself is eminently open-sourceable, and having a community will help us keep pace with fast-moving Docker changes, as well as opening up a plugin ecosystem.

Migration roadmap

Migration begins with coreservices (midtier Thrift services) and their respective persistence layers. No more downloading, compiling, and installing Cassandra, MySQL, Redis, Memcache, etc. Thrift is now automatically compiled at import time, using the newest version of Thrift. Once coreservices is proven, we move onto Pylons apps. Once that has momentum, we can shift our attention to refactoring and revamping lib.