Automating releasing code changes, then staging them in a continuous delivery fashion and in an easy-to-visualize fashion with AWS CodePipeline was really big, as was unifying our deployment patterns with AWS CodeDeploy. 
Brian Hare Engineering Lead for cloud platform development

CareerBuilder operates one of the largest online job sites in the US and is a leader in human capital solutions. It provides employers with tools for job distribution, sourcing, workflow, CRM, data, and analytics in a single platform. With more than 24 million monthly unique visitors and relationships with employers that include 92 percent of the Fortune 1000, CareerBuilder helps connect job seekers with employment opportunities. The company powers career sites for more than 1,000 partners and operates in the United States, Europe, Canada, Asia, and South America.

  • CareerBuilder’s central API hub is a growing platform that handles all traffic and routing requests for every CareerBuilder service.
  • With 200M requests daily, the hub has high throughput and low tolerance for failure.
  • The engineering team was using a combination of Jenkins, Chef, and manual deployment scripts to update applications and configurations.
  • Deployments were manual and lengthy, and thus performed irregularly.
  • Deployment scripts broke whenever Jenkins versions were updated, bugs would accidentally get released, and configuration updates had to be hardcoded into every application update.
  • The team needed a simpler way to deploy updates to the hub and lower deployment risk.
  • CareerBuilder now uses AWS CodePipeline to model and then automate the software release process.
  • Whenever a change is made, a staging pipeline (“Pipeline A” in the illustration), picks up application code changes from GitHub, triggers unit and regression tests in Jenkins, and then deploys to a staging environment using AWS CodeDeploy.
  • The same pipeline also detects changes to configuration files stored in Amazon Simple Storage Service (Amazon S3) and automatically moves the changes to the staging environment using a parallel stage.
  • Two identical production pipelines (pipelines “B” and “C” in the illustration) allow CareerBuilder to more easily perform blue-green deployments; each pipeline is disabled by default and then activated by the team depending on which instances they want to update.
  • When the team is ready to deploy to production, AWS CodePipeline triggers AWS CodeDeploy to deploy the application updates and configuration changes to Amazon Elastic Compute Cloud (Amazon EC2); they no longer need Jenkins, Chef server, or manual deployment scripts to deploy to production.
careerbuilder-arch-diagram

The illustration above displays the following steps: 1) CodePipeline detects changes to application and configuration code. 2) CodePipeline then triggers automated build and test in Jenkins. 3) CodePipeline uses CodeDeploy to deploy the output build artifact and configuration changes to staging. 4) CareerBuilder leverages CodePipeline to implement a blue-green deployment. CodePipeline (Pipeline C) is enabled when CareerBuilder is ready to deploy the update to new production instances.

 

  • Continuous delivery: CareerBuilder can integrate, test, and stage changes to applications quickly and automatically, enabling faster time to market and increased agility.
  • Quality assurance: AWS CodePipeline ensures that every code change automatically runs through the integration tests that the team defines.
  • More efficient use of resources: CareerBuilder eliminated the need to run and manage a Chef server for configuration updates, freeing up an entire team to focus on development work.
  • Improved productivity: Engineers can deploy their changes faster and more frequently because CodePipeline and CodeDeploy automate many manual aspects and steps of deployment.

To learn more about how AWS can help enable continuous delivery and automate software deployments, visit our AWS CodePipeline and AWS CodeDeploy pages.