OpenSearch
Apache-2.0 fork of Elasticsearch 7.10, governed by the OpenSearch Software Foundation with the same core architecture.
What it is
A fork of Elasticsearch 7.10 and Kibana, created in 2021 after Elastic changed its licence. It is Apache-2.0 and, since 2024, governed by the OpenSearch Software Foundation under the Linux Foundation.
Architecture
Identical in origin: Lucene-backed shards, replicas, immutable segments with background merges, and separated node roles. The entire Elasticsearch and OpenSearch documentation section applies to both, with per-page notes where API paths differ.
Best use cases
The same as Elasticsearch — full-text search, log and observability search, faceted navigation — with two additional considerations:
- Licensing. Apache-2.0 matters for vendors redistributing the software and for organisations with policies against source-available licences.
- Bundled security features. Authentication, role-based access control, encryption and audit logging are included rather than gated behind a commercial tier.
When not to use it
The same exclusions as Elasticsearch: not a system of record, poor fit for frequently updated documents, limited joins.
Additionally: if you depend on an Elastic-specific feature or an integration that targets recent Elasticsearch versions only, verify support before committing.
Data model
The same JSON documents, mappings, analysers and the text versus keyword distinction. Existing
index templates and mappings generally transfer.
Consistency and transactions
The same model: no cross-document transactions, refresh-based visibility, translog durability settings.
Scaling model
The same: shards fixed at index creation, replicas adjustable, tiered node roles for data age. Index State Management (ISM) fills the role of Elasticsearch's ILM, with different API paths and policy syntax.
Replication
The same primary and replica shard model, with the same cluster health semantics.
Backup and recovery
Snapshots to a repository, with the same segment-level incremental behaviour and the same version-compatibility constraints on restore.
Monitoring
The same _cluster/health, _cat APIs and node stats. Metric collection and dashboards generally
transfer, but exporters and integrations should be verified against OpenSearch specifically.
Common mistakes
The Elasticsearch mistakes apply unchanged — excessive shard counts, oversized heaps, dynamic mapping on external documents, deep pagination.
Production checklist
The search production checklist applies to OpenSearch, with ISM in place of ILM.