This document discusses replicating Elasticsearch clusters across multiple datacenters at sahibinden.com to prevent downtime, enable backups and testing, and provide low latency access. It describes three options for replication: synchronous writes, asynchronous writes, and a hybrid approach. Sahibinden.com uses Apache Kafka and Zookeeper for asynchronous cross-cluster and cross-datacenter replication between four Elasticsearch clusters in two datacenters. Performance is maintained using multiple replication threads. A sync script coordinates the replication process.