Dromedary...without servers.
JavaScript
Permalink
Failed to load latest commit information.
docs update pipeline image Mar 16, 2016
pipeline update HOME to '/tmp' to allow newer version of npm write cache to fi… Oct 28, 2016
.gitignore crop images Feb 23, 2016
LICENSE.md add license Aug 10, 2016
README.md Update README.md Apr 18, 2016
gulpfile.js revert default hosts May 3, 2016
package.json

README.md

Overview

This project deploys dromedary in AWS Lambda with API Gateway as the interface to demonstrate serverless architecture. It also demonstrates the use of CodePipeline with Lambdas to continuously deliver changes made in the source code in a serverless manner. We've also written a three-part blog series on the topic here: Serverless Delivery: Architecture (Part 1), Serverless Delivery: Bootstrapping the Pipeline (Part 2) and Serverless Delivery: Orchestrating the Pipeline (Part 3).

Architecture Overview

The application is split into 2 separate parts for deployment:

  • API - deployed as a Lambda function using API Gateway for the front end.
  • Static Content - deployed into an S3 bucket with website hosting enabled.

Additionally, a config.json file is generated and deployed into the S3 bucket containing the endpoint to use for the API in API Gateway.

app-overview

Pipeline Overview

The pipeline consists of the following steps:

  • commit - a commit in GitHub triggers a new CodePipeline job. The source is downloaded from GitHub and then pushed into an S3 bucket as a zip file.
  • npm lambda - a lambda is executed that downloads the source zip file, runs npm install to get he dependencies and then uploads the source+dependencies to S3 as a tarball.
  • gulp lambda(s) - a lambda is executed that downloads the source+dependencies tarball from S3, extracts it, then runs a gulp task

The details of what happens in the gulp task is completely owned by the gulpfile.js in the source code. This provides decoupling of the pipeline from the app and allows the pipeline template to be used by any gulp project.

pipeline-overview

Here's a sample of what the pipeline looks like in AWS CodePipeline console:

pipeline-example

Launching Pipeline

To integrate with GitHub, AWS CodePipeline uses OAuth tokens. Generate your token at GitHub and ensure you enable the following two scopes:

  • admin:repo_hook, which is used to detect when you have committed and pushed changes to the repository
  • repo, which is used to read and pull artifacts from public and private repositories into a pipeline

You can launch via the console: Launch Pipeline stack

Or you can launch by using gulp in this repo:

  • PREREQUISITES - You need Node.js installed.
    • Install Node.js: sudo yum install nodejs npm --enablerepo=epel (For OS X, check out nodejs.org)
    • Update NPM: curl -L https://npmjs.org/install.sh | sudo sh
    • Install Gulp: sudo npm install -g gulp
  • Download this repo and then run npm install first to install all dependent modules.
  • Bring the pipeline up with gulp pipeline:up --token=XXXXXXXXXXXXXXXXX
  • You can run gulp pipeline:wait to wait for the stack to come up, and then gulp pipeline:status to get the outputs and gulp pipeline:stacks to see what applicaiton stacks the pipeline has currently running.
  • To tear everything down, run gulp pipeline:teardown
  • By default, the stack name will be dromedary-serverless. You can change this by passing --stackName=my-stack-name to any of the above gulp commands.

Development

To publish pipeline template/lambda changes, run gulp publish. You may want to choose a different bucket to publish the templates and lambda code to via the --templateBucket argument.

Todo

  • Tighten up IAM policies in CFN