Indexing process

1. When you index a document, the node that receives your HTTP request determines which shard should contain your document, based on a hash of the document’s ID.
2. Then, it forwards the document to the node holding that primary shard.
3. Once the shard is updated, the document is sent to be indexed in all replicas of that shard.
4. Finally, the indexing command returns after all replicas are updated.

This process keeps replicas in sync and allows them to serve searches and to be promoted to primary shards in case the original primary is unavailable.

Searching an index requires Elasticsearch to query a set of shards that contain all your data. Since shards are distributed across multiple nodes, this query will be forwarded to multiple nodes.

Search process

1. When you make a search request, the node that receives it forwards it to a set of shards that contain all your data. These can be either primary shards or replicas, and are chosen in a round-robin fashion.
2. Then, the coordinator node gathers the results from those shards, aggregates and sorts them, and sends the results back to the client.