Tuesday 29th September, 2015
2:30pm to 3:20pm
Traditional infrastructures closely couple hostnames to machine roles. It's desirable to be able to trivially scale clusters up or down, and as you scale up frequent machine replacement becomes a fact of life.
Therefore, the need to pre-allocate DNS for individual machines becomes extremely limiting. This talk will explain how the Yelp operations team uses puppet for configuration management - driven by YAML data and EC2 tags, with Terraform taking care of orchestration above the single machine level, to build out generic clusters of common components.
Using Packer to bake common base AMIs, and well defined puppet 'roles', and flexible service discovery (via Smartstack) and monitoring (via Sensu) has empowered individual engineering teams to manage and own their own infrastructure.
Teams can launch, manage and monitor their own Elasticsearch and Cassandra clusters in multiple AWS regions without the operations team's involvement - whilst also taking advantage of organization wide best practices and monitoring.
Site Reliability Engineer at Yelp
Sign in to add slides, notes or videos to this session