Lucene/Solr Revolution 2017

Home Depot's online search platform which powers our ecommerce platform requires an engine that can handle high volume reads, real time indexing and batch updates on nodes taking live traffic. Leveraging Solr Cloud's distributed architecture and google cloud infrastructure, we were able to provision an HA and scalable Solr cluster across multi regions to support the high volume reads and indexing.

This session will focus on: the provisioning and automation of building a Solr cluster in the cloud; challenges involved in automation of Solr cluster creations on cloud platforms like Google Cloud Platform and how our team overcame them; replica management in ephemeral cloud infrastructure; how to plug Solr into Google Cloud Instance Group infrastructure; and learnings from usage of Go language plugins for Zookeeper and in house Go language collection management package for Solr.

We'll also cover patterns on routing traffic to a newly provisioned cluster without losing real time indexing feeds to support version upgrades or any changes. We'll talk about our performance learnings and optimizations which helped meet our aggressive SLOs. In the end we'll also cover the metrics, alerting and automated recovery patterns we implemented for Solr, Zookeeper and Fusion stack to ensure reliability.