Low-level clients#

Clients provide a low-level interface to AWS whose methods map close to 1:1 with service APIs. All service operations are supported by clients. Clients are generated from a JSON service definition file.

Creating clients#

Clients are created in a similar fashion to resources:

import boto3

# Create a low-level client with the service name
sqs = boto3.client('sqs')

It is also possible to access the low-level client from an existing resource:

# Create the resource
sqs_resource = boto3.resource('sqs')

# Get the client from the resource
sqs = sqs_resource.meta.client

Service operations#

Service operations map to client methods of the same name and provide access to the same operation parameters via keyword arguments:

# Make a call using the low-level client
response = sqs.send_message(QueueUrl='...', MessageBody='...')

As can be seen above, the method arguments map directly to the associated SQS API.


The method names have been snake-cased for better looking Python code.

Parameters must be sent as keyword arguments. They will not work as positional arguments.

Handling responses#

Responses are returned as python dictionaries. It is up to you to traverse or otherwise process the response for the data you need, keeping in mind that responses may not always include all of the expected data. In the example below, response.get('QueueUrls', []) is used to ensure that a list is always returned, even when the response has no key 'QueueUrls':

# List all your queues
response = sqs.list_queues()
for url in response.get('QueueUrls', []):

The response in the example above looks something like this:

    "QueueUrls": [


Waiters use a client’s service operations to poll the status of an AWS resource and suspend execution until the AWS resource reaches the state that the waiter is polling for or a failure occurs while polling. Using clients, you can learn the name of each waiter that a client has access to:

import boto3

s3 = boto3.client('s3')
sqs = boto3.client('sqs')

# List all of the possible waiters for both clients
print("s3 waiters:")

print("sqs waiters:")

Note if a client does not have any waiters, it will return an empty list when accessing its waiter_names attribute:

s3 waiters:
[u'bucket_exists', u'bucket_not_exists', u'object_exists', u'object_not_exists']
sqs waiters:

Using a client’s get_waiter() method, you can obtain a specific waiter from its list of possible waiters:

# Retrieve waiter instance that will wait till a specified
# S3 bucket exists
s3_bucket_exists_waiter = s3.get_waiter('bucket_exists')

Then to actually start waiting, you must call the waiter’s wait() method with the method’s appropriate parameters passed in:

# Begin waiting for the S3 bucket, mybucket, to exist

Multithreading or multiprocessing with clients#

Unlike Resources and Sessions, clients are generally thread-safe. There are some caveats, defined below, to be aware of though.


Multi-Processing: While clients are thread-safe, they cannot be shared across processes due to their networking implementation. Doing so may lead to incorrect response ordering when calling services.

Shared Metadata: Clients expose metadata to the end user through a few attributes (namely meta, exceptions and waiter_names). These are safe to read but any mutations should not be considered thread-safe.

Custom Botocore Events: Botocore (the library Boto3 is built on) allows advanced users to provide their own custom event hooks which may interact with boto3’s client. The majority of users will not need to use these interfaces, but those that do should no longer consider their clients thread-safe without careful review.


boto3.client('<service_name>') is an alias for creating a client with a shared default session. Invoking boto3.client() inside of a concurrent context may result in response ordering issues or interpreter failures from underlying SSL modules.

General Example#

import boto3.session
from concurrent.futures import ThreadPoolExecutor

def do_s3_task(client, task_definition):
    # Put your thread-safe code here

def my_workflow():
    # Create a session and use it to make our client
    session = boto3.session.Session()
    s3_client = session.client('s3')

    # Define some work to be done, this can be anything
    my_tasks = [ ... ]

    # Dispatch work tasks with our s3_client
    with ThreadPoolExecutor(max_workers=8) as executor:
        futures = [executor.submit(do_s3_task, s3_client, task) for task in my_tasks]