Async AWS SDK for Python

Breaking changes for v11: The S3Transfer config passed into upload/download_file etc.. has been updated to that it matches what boto3 uses

Breaking changes for v9: aioboto3.resource and aioboto3.client methods no longer exist, make a session then call session.client etc... This was done for various reasons but mainly that it prevents the default session living longer than it should as that breaks situations where eventloops are replaced.

The .client and .resource functions must now be used as async context managers.

Now that aiobotocore has reached version 1.0.1, a side effect of the work put in to fix various issues like bucket region redirection and supporting web assume role type credentials, the client must now be instantiated using a context manager, which by extension applies to the resource creator. You used to get away with calling res = aioboto3.resource('dynamodb') but that no longer works. If you really want to do that, you can do res = await aioboto3.resource('dynamodb').__aenter__() but you'll need to remember to call __aexit__.

There will most likely be some parts that dont work now which I've missed, just make an issue and we'll get them resoved quickly.

Creating service resources must also be async now, e.g.

async def main():
    session = aioboto3.Session()
    async with session.resource("s3") as s3:
        bucket = await s3.Bucket('mybucket')  # <----------------
        async for s3_object in bucket.objects.all():
            print(s3_object)

Updating to aiobotocore 1.0.1 also brings with it support for running inside EKS as well as asyncifying get_presigned_url

This package is mostly just a wrapper combining the great work of boto3 and aiobotocore.

aiobotocore allows you to use near enough all of the boto3 client commands in an async manner just by prefixing the command with await.

With aioboto3 you can now use the higher level APIs provided by boto3 in an asynchronous manner. Mainly I developed this as I wanted to use the boto3 dynamodb Table object in some async microservices.

While all resources in boto3 should work I havent tested them all, so if what your after is not in the table below then try it out, if it works drop me an issue with a simple test case and I'll add it to the table.

Example

Simple example of using aioboto3 to put items into a dynamodb table

import asyncio
import aioboto3
from boto3.dynamodb.conditions import Key


async def main():
    session = aioboto3.Session()
    async with session.resource('dynamodb', region_name='eu-central-1') as dynamo_resource:
        table = await dynamo_resource.Table('test_table')

        await table.put_item(
            Item={'pk': 'test1', 'col1': 'some_data'}
        )

        result = await table.query(
            KeyConditionExpression=Key('pk').eq('test1')
        )

        # Example batch write
        more_items = [{'pk': 't2', 'col1': 'c1'}, \
                      {'pk': 't3', 'col1': 'c3'}]
        async with table.batch_writer() as batch:
            for item_ in more_items:
                await batch.put_item(Item=item_)

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

# Outputs:
#  [{'col1': 'some_data', 'pk': 'test1'}]

Things that either dont work or have been patched

As this library literally wraps boto3, its inevitable that some things won't magically be async.

Fixed:

• s3_client.download_file* This is performed by the s3transfer module. -- Patched with get_object
• s3_client.upload_file* This is performed by the s3transfer module. -- Patched with custom multipart upload
• s3_client.copy This is performed by the s3transfer module. -- Patched to use get_object -> upload_fileobject
• dynamodb_resource.Table.batch_writer This now returns an async context manager which performs the same function
• Resource waiters - You can now await waiters which are part of resource objects, not just client waiters, e.g. await dynamodbtable.wait_until_exists()
• Resource object properties are normally autoloaded, now they are all co-routines and the metadata they come from will be loaded on first await and then cached thereafter.
• S3 Bucket.objects object now works and has been asyncified. Examples here - https://aioboto3.readthedocs.io/en/latest/usage.html#s3-resource-objects

Amazon S3 Client-Side Encryption

Boto3 doesn't support AWS client-side encryption so until they do I've added basic support for it. Docs here CSE

CSE requires the python cryptography library so if you do pip install aioboto3[s3cse] that'll also include cryptography.

This library currently supports client-side encryption using KMS-Managed master keys performing envelope encryption using either AES/CBC/PKCS5Padding or preferably AES/GCM/NoPadding. The files generated are compatible with the Java Encryption SDK so I will assume they are compatible with the Ruby, PHP, Go and C++ libraries as well.

Non-KMS managed keys are not yet supported but if you have use of that, raise an issue and i'll look into it.

Documentation

Docs are here - https://aioboto3.readthedocs.io/en/latest/

Examples here - https://aioboto3.readthedocs.io/en/latest/usage.html

Features

• Closely mimics the usage of boto3.

Todo

• More examples
• Set up docs
• Look into monkey-patching the aws xray sdk to be more async if it needs to be.

Credits

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template. It also makes use of the aiobotocore and boto3 libraries. All the credit goes to them, this is mainly a wrapper with some examples.