TLDR Tech / Data
How we rebuilt the search architecture for high availability in GitHub Enterprise Server
GitHub rebuilt its search architecture using Elasticsearch's Cross Cluster Replication (CCR) to run independent single-node clusters per instance (primary and replicas). This approach enables durable persistence, asynchronous replication after Lucene segments are created, custom workflows for setup and failover, zero-downtime migrations, and automatic replica promotion for failover.
Read the full story