You're viewing version 2.17 of the OpenSearch documentation. This version is no longer maintained. For the latest version, see the current documentation. For information about OpenSearch version maintenance, see Release Schedule and Maintenance Policy.
Streaming bulk
Introduced 2.17.0
This is an experimental feature and is not recommended for use in a production environment. For updates on the progress of the feature or if you want to leave feedback, see the associated GitHub issue.
The streaming bulk operation lets you add, update, or delete multiple documents by streaming the request and getting the results as a streaming response. In comparison to the traditional Bulk API, streaming ingestion eliminates the need to estimate the batch size (which is affected by the cluster operational state at any given time) and naturally applies backpressure between many clients and the cluster. The streaming works over HTTP/2 or HTTP/1.1 (using chunked transfer encoding), depending on the capabilities of the clients and the cluster.
The default HTTP transport method does not support streaming. You must install the transport-reactor-netty4 HTTP transport plugin and use it as the default HTTP transport layer. Both the transport-reactor-netty4 plugin and the Streaming Bulk API are experimental.
Path and HTTP methods
POST _bulk/stream
POST <index>/_bulk/stream
If you specify the index in the path, then you don’t need to include it in the request body chunks.
OpenSearch also accepts PUT requests to the _bulk/stream path, but we highly recommend using POST. The accepted usage of PUT—adding or replacing a single resource on a given path—doesn’t make sense for streaming bulk requests.
Query parameters
The following table lists the available query parameters. All query parameters are optional.
| Parameter | Data type | Description |
|---|---|---|
pipeline | String | The pipeline ID for preprocessing documents. |
refresh | Enum | Whether to refresh the affected shards after performing the indexing operations. Default is false. true causes the changes show up in search results immediately but degrades cluster performance. wait_for waits for a refresh. Requests take longer to return, but cluster performance isn’t degraded. |
require_alias | Boolean | Set to true to require that all actions target an index alias rather than an index. Default is false. |
routing | String | Routes the request to the specified shard. |
timeout | Time | How long to wait for the request to return. Default is 1m. |
type | String | (Deprecated) The default document type for documents that don’t specify a type. Default is _doc. We highly recommend ignoring this parameter and using the _doc type for all indexes. |
wait_for_active_shards | String | Specifies the number of active shards that must be available before OpenSearch processes the bulk request. Default is 1 (only the primary shard). Set to all or a positive integer. Values greater than 1 require replicas. For example, if you specify a value of 3, the index must have 2 replicas distributed across 2 additional nodes in order for the request to succeed. |
batch_interval | Time | Specifies for how long bulk operations should be accumulated into a batch before sending the batch to data nodes. |
batch_size | Time | Specifies how many bulk operations should be accumulated into a batch before sending the batch to data nodes. Default is 1. |
Request body fields
The Streaming Bulk API request body is fully compatible with the Bulk API request body, where each bulk operation (create/index/update/delete) is sent as a separate chunk.
Example request
curl -X POST "http://localhost:9200/_bulk/stream" -H "Transfer-Encoding: chunked" -H "Content-Type: application/json" -d'
{ "delete": { "_index": "movies", "_id": "tt2229499" } }
{ "index": { "_index": "movies", "_id": "tt1979320" } }
{ "title": "Rush", "year": 2013 }
{ "create": { "_index": "movies", "_id": "tt1392214" } }
{ "title": "Prisoners", "year": 2013 }
{ "update": { "_index": "movies", "_id": "tt0816711" } }
{ "doc" : { "title": "World War Z" } }
'
Example response
Depending on the batch settings, each streamed response chunk may report the results of one or many (batch) bulk operations. For example, for the preceding request with no batching (default), the streaming response may appear as follows:
{"took": 11, "errors": false, "items": [ { "index": {"_index": "movies", "_id": "tt1979320", "_version": 1, "result": "created", "_shards": { "total": 2 "successful": 1, "failed": 0 }, "_seq_no": 1, "_primary_term": 1, "status": 201 } } ] }
{"took": 2, "errors": true, "items": [ { "create": { "_index": "movies", "_id": "tt1392214", "status": 409, "error": { "type": "version_conflict_engine_exception", "reason": "[tt1392214]: version conflict, document already exists (current version [1])", "index": "movies", "shard": "0", "index_uuid": "yhizhusbSWmP0G7OJnmcLg" } } } ] }
{"took": 4, "errors": true, "items": [ { "update": { "_index": "movies", "_id": "tt0816711", "status": 404, "error": { "type": "document_missing_exception", "reason": "[_doc][tt0816711]: document missing", "index": "movies", "shard": "0", "index_uuid": "yhizhusbSWmP0G7OJnmcLg" } } } ] }