ElasticCC: Building your Elasticsearch cloud on AWS using EC2, SSM and Lambda

With Andrej Blagojevic
Intercom is one of the world’s leading conversational support platforms. We are used by over 30000 companies worldwide, including customers like Facebook, Amazon, Shopify, Sotheby’s and more. At Intercom, Elasticsearch forms one of the core datastore technologies and is our choice for powering full-text search and real-time reporting features available to customers, as well as powering our conversational support chatbots and more using Elasticsearch machine learning features. Our clusters contain hundreds of nodes, store more than 20 billion documents and serve around 12 million queries every minute.
They also predate technologies of containerization, requiring tight integration with some of our core customer workflows alongside the ability to really finely tune the clusters for maximum performance. And so from day one, as a company solely running on AWS cloud we went on a journey to design and implement our own Elastic cloud on top of EC2 allowing our product teams straightforward ways to provision, monitor and seamlessly maintain their Elasticsearch infrastructure, including automated security patching, version upgrades, vertical and horizontal cluster scaling and more. This is a story of the stack we have built and the lessons we have learned along the way.