Open Source Projects · Sub-chapter 3 · 4 min read

OpenSearch

Sub-chapter 3 of Open Source Projects · Search, analytics, and vectors - under one Apache 2.0 roof

OpenSearch is one of the most consequential forks in modern open source. When Elastic relicensed Elasticsearch and Kibana to the non-OSI SSPL/Elastic License v2 in 2021, AWS forked the last Apache 2.0 release (Elasticsearch 7.10.2) and made the result a community-owned project. In September 2024 AWS handed it to the Linux Foundation as the neutral OpenSearch Software Foundation.

What makes it interesting for a fresh grad: OpenSearch is large enough to be career-shaping to contribute to, small enough that the maintainers actually read your PR, and on a topic - search + analytics + vectors - that touches every Welzin product.

TL;DR

What - Distributed search, analytics, and vector-database engine. Fork of Elasticsearch 7.10 + Kibana.
License - Apache 2.0 (the whole point of the fork).
Governance - OpenSearch Software Foundation under the Linux Foundation since Sep 16, 2024.
Components - OpenSearch (engine) + OpenSearch Dashboards (the Kibana fork).
Use cases - full-text search, log analytics, observability, semantic / vector search, RAG.

Why it exists

In January 2021, Elastic relicensed Elasticsearch and Kibana to the Server Side Public License (SSPL) and the Elastic License v2 - both non-OSI-approved. The change effectively prevented AWS and other cloud vendors from offering Elasticsearch as a managed service without paying Elastic. AWS responded by forking the last Apache-2.0 version (Elasticsearch 7.10.2 / Kibana 7.10.2) and renaming the result OpenSearch and OpenSearch Dashboards.

For three years, OpenSearch was AWS-controlled. In September 2024, AWS transferred ownership to a new neutral entity under the Linux Foundation - the OpenSearch Software Foundation - with SAP and Uber as Premier members alongside AWS, plus Aiven, Atlassian, Canonical, DigitalOcean, NetApp, and others. The handoff resolved the "is OpenSearch really open?" question definitively.

Adoption: 700M+ downloads, used in production at AWS, SAP, Uber, Stripe, Adobe, and thousands of smaller companies.

Architecture - the mental model

OpenSearch is a JVM-based distributed system built on Apache Lucene (the same indexing library Elasticsearch and Solr use).

text

   Documents → Index → Shards (primary + replica)
                          ↓
                       Nodes
                          ↓
                       Cluster
                          ↓
            REST/JSON API ←→ Client / Dashboards

Key concepts:

Index - a logical collection of documents (think: a database table).
Shard - index data is split into primary shards (for scale) and replica shards (for HA).
Node - a JVM running OpenSearch.
Cluster - a group of nodes that coordinate via the master node.

Plugins add capabilities: Security, Alerting, k-NN (vector search), ML Commons (model serving), Index Management, Anomaly Detection, Observability, SQL/PPL, Notifications.

OpenSearch Dashboards is the UI layer - the Kibana fork - for visualization, exploration, and managing the cluster.

For RAG / AI workloads: the k-NN plugin stores dense vector embeddings and runs approximate nearest-neighbour search. OpenSearch is one of the few major search engines where lexical (BM25) and vector retrieval coexist natively - perfect for hybrid search.

License and governance

License: Apache 2.0
Foundation: OpenSearch Software Foundation (under Linux Foundation, Sep 2024)
Premier members: AWS, SAP, Uber

Install and run locally

The fastest path is Docker:

bash

docker run -d --name opensearch \
  -p 9200:9200 -p 9600:9600 \
  -e "discovery.type=single-node" \
  -e "OPENSEARCH_INITIAL_ADMIN_PASSWORD=MyStrongP@ss1" \
  opensearchproject/opensearch:latest

Then talk to it:

bash

# Health check
curl -ku admin:MyStrongP@ss1 https://localhost:9200/_cluster/health

# Index a document
curl -ku admin:MyStrongP@ss1 -X PUT https://localhost:9200/welzin/_doc/1 \
  -H 'Content-Type: application/json' \
  -d '{"title": "Hello OpenSearch", "tags": ["bootcamp"]}'

# Search
curl -ku admin:MyStrongP@ss1 https://localhost:9200/welzin/_search?q=hello

For the UI, add Dashboards:

bash

docker run -d --name opensearch-dashboards \
  --link opensearch \
  -p 5601:5601 \
  -e "OPENSEARCH_HOSTS=[\"https://opensearch:9200\"]" \
  opensearchproject/opensearch-dashboards:latest

Open http://localhost:5601.

The official getting-started tutorial: docs.opensearch.org/latest/getting-started.

How to contribute

GitHub org: github.com/opensearch-project - dozens of repos.
Community repo + contributor guide: github.com/opensearch-project/community.
Onboarding doc: .github/ONBOARDING.md.
Good first issue: filter across the org with this search.
Slack: opensearch.org/slack.
Forum: forum.opensearch.org.

Which repo to start on:

opensearch-project/OpenSearch - the engine itself (Java). High bar, large codebase.
opensearch-project/OpenSearch-Dashboards - the UI (TypeScript/React). More approachable for web devs.
opensearch-project/documentation-website - docs (Markdown). The friendliest entry-point.
opensearch-project/ml-commons - ML plugin (Java). Interesting if you're going through Chapter VI.

Hands-on Checkpoints

Run OpenSearch + Dashboards locally via Docker. Index 20 documents (e.g., the chapters from this bootcamp). Search them.
Enable the k-NN plugin. Index dense-vector embeddings of those documents. Query by similarity.
Browse opensearch-project/documentation-website issues. Pick one good first issue doc fix.
Read the OpenSearch RFC process. Understand how new features are proposed.
Join the OpenSearch Slack. Lurk in #core for a week.