Dev team open-sources SeaSearch, an S3-backed Elasticsearch alternative for multi-tenant search

A development team has open-sourced SeaSearch, an Elasticsearch-compatible search engine built to overcome the scaling limits of running one index per tenant on Elasticsearch. Rather than replicating data across nodes, SeaSearch stores all index data in S3-compatible object storage and routes query ownership via a partition map held in etcd. The system is built on ZincSearch and the Bluge indexing library, with compute nodes holding no authoritative data, making failover a metadata update instead of a data migration. To address the latency penalty of object storage, SeaSearch uses parallel segment warm-up and distributed query execution across multiple compute nodes. The team notes the architecture suits workloads with a small active data subset, and has been running in production for two years, though it omits some Elasticsearch features such as shard and replica settings.