Using Asyncio with Elasticsearch

Starting in elasticsearch-py v7.8.0 for Python 3.6+ the elasticsearch package supports async/await with Asyncio and Aiohttp. You can either install aiohttp directly or use the [async] extra:

$ python -m pip install elasticsearch>=7.8.0 aiohttp

# - OR -

$ python -m pip install elasticsearch[async]>=7.8.0

Note

Async functionality is a new feature of this library in v7.8.0+ so please open an issue if you find an issue or have a question about async support.

Getting Started with Async

After installation all async API endpoints are available via AsyncElasticsearch and are used in the same way as other APIs, just with an extra await:

import asyncio
from elasticsearch import AsyncElasticsearch

es = AsyncElasticsearch()

async def main():
    resp = await es.search(
        index="documents",
        query={"match_all": {}},
        size=20,
    )
    print(resp)

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

All APIs that are available under the sync client are also available under the async client.

ASGI Applications and Elastic APM

ASGI (Asynchronous Server Gateway Interface) is a new way to serve Python web applications making use of async I/O to achieve better performance. Some examples of ASGI frameworks include FastAPI, Django 3.0+, and Starlette. If you’re using one of these frameworks along with Elasticsearch then you should be using AsyncElasticsearch to avoid blocking the event loop with synchronous network calls for optimal performance.

Elastic APM also supports tracing of async Elasticsearch queries just the same as synchronous queries. For an example on how to configure AsyncElasticsearch with a popular ASGI framework FastAPI and APM tracing there is a pre-built example in the examples/fastapi-apm directory.

Frequently Asked Questions

NameError / ImportError when importing AsyncElasticsearch?

If when trying to use AsyncElasticsearch and you’re receiving a NameError or ImportError you should ensure that you’re running Python 3.6+ (check with $ python --version) and that you have aiohttp installed in your environment (check with $ python -m pip freeze | grep aiohttp). If either of the above conditions is not met then async support won’t be available.

What about the elasticsearch-async package?

Previously asyncio was supported separately via the elasticsearch-async package. The elasticsearch-async package has been deprecated in favor of AsyncElasticsearch provided by the elasticsearch package in v7.8 and onwards.

Receiving ‘Unclosed client session / connector’ warning?

This warning is created by aiohttp when an open HTTP connection is garbage collected. You’ll typically run into this when closing your application. To resolve the issue ensure that close() is called before the AsyncElasticsearch instance is garbage collected.

For example if using FastAPI that might look like this:

from fastapi import FastAPI
from elasticsearch import AsyncElasticsearch

app = FastAPI()
es = AsyncElasticsearch()

# This gets called once the app is shutting down.
@app.on_event("shutdown")
async def app_shutdown():
    await es.close()

Async Helpers

Async variants of all helpers are available in elasticsearch.helpers and are all prefixed with async_*. You’ll notice that these APIs are identical to the ones in the sync Helpers documentation.

All async helpers that accept an iterator or generator also accept async iterators and async generators.

Bulk and Streaming Bulk

import asyncio
from elasticsearch import AsyncElasticsearch
from elasticsearch.helpers import async_bulk

es = AsyncElasticsearch()

async def gendata():
    mywords = ['foo', 'bar', 'baz']
    for word in mywords:
        yield {
            "_index": "mywords",
            "doc": {"word": word},
        }

async def main():
    await async_bulk(es, gendata())

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

import asyncio
from elasticsearch import AsyncElasticsearch
from elasticsearch.helpers import async_streaming_bulk

es = AsyncElasticsearch()

async def gendata():
    mywords = ['foo', 'bar', 'baz']
    for word in mywords:
        yield {
            "_index": "mywords",
            "word": word,
        }

async def main():
    async for ok, result in async_streaming_bulk(es, gendata()):
        action, result = result.popitem()
        if not ok:
            print("failed to %s document %s" % ())

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

Scan

import asyncio
from elasticsearch import AsyncElasticsearch
from elasticsearch.helpers import async_scan

es = AsyncElasticsearch()

async def main():
    async for doc in async_scan(
        client=es,
        query={"query": {"match": {"title": "python"}}},
        index="orders-*"
    ):
        print(doc)

loop = asyncio.get_event_loop()
loop.run_until_complete(main())

Reindex

API Reference

The API of AsyncElasticsearch is nearly identical to the API of Elasticsearch with the exception that every API call like search() is an async function and requires an await to properly return the response body.

AsyncElasticsearch

Note

To reference Elasticsearch APIs that are namespaced like .indices.create() refer to the sync API reference. These APIs are identical between sync and async.

AsyncTransport

AsyncConnection

AIOHttpConnection