The AWS Python SDK team does not intend to add new features to the resources interface in boto3. Existing interfaces will continue to operate during boto3’s lifecycle. Customers can find access to newer service features through the client interface.

Resources represent an object-oriented interface to Amazon Web Services (AWS). They provide a higher-level abstraction than the raw, low-level calls made by service clients. To use resources, you invoke the resource() method of a Session and pass in a service name:

# Get resources from the default session
sqs = boto3.resource('sqs')
s3 = boto3.resource('s3')

Every resource instance has a number of attributes and methods. These can conceptually be split up into identifiers, attributes, actions, references, sub-resources, and collections. Each of these is described in further detail below and in the following section.

Resources themselves can also be conceptually split into service resources (like sqs, s3, ec2, etc) and individual resources (like sqs.Queue or s3.Bucket). Service resources do not have identifiers or attributes. The two share the same components otherwise.

Identifiers and attributes#

An identifier is a unique value that is used to call actions on the resource. Resources must have at least one identifier, except for the top-level service resources (e.g. sqs or s3). An identifier is set at instance creation-time, and failing to provide all necessary identifiers during instantiation will result in an exception. Examples of identifiers:

# SQS Queue (url is an identifier)
queue = sqs.Queue(url='http://...')

# S3 Object (bucket_name and key are identifiers)
obj = s3.Object(bucket_name='boto3', key='')

# Raises exception, missing identifier: key!
obj = s3.Object(bucket_name='boto3')

Identifiers may also be passed as positional arguments:

# SQS Queue
queue = sqs.Queue('http://...')

# S3 Object
obj = s3.Object('boto3', '')

# Raises exception, missing key!
obj = s3.Object('boto3')

Identifiers also play a role in resource instance equality. For two instances of a resource to be considered equal, their identifiers must be equal:

>>> bucket1 = s3.Bucket('boto3')
>>> bucket2 = s3.Bucket('boto3')
>>> bucket3 = s3.Bucket('some-other-bucket')

>>> bucket1 == bucket2
>>> bucket1 == bucket3


Only identifiers are taken into account for instance equality. Region, account ID and other data members are not considered. When using temporary credentials or multiple regions in your code please keep this in mind.

Resources may also have attributes, which are lazy-loaded properties on the instance. They may be set at creation time from the response of an action on another resource, or they may be set when accessed or via an explicit call to the load or reload action. Examples of attributes:

# SQS Message

# S3 Object


Attributes may incur a load action when first accessed. If latency is a concern, then manually calling load will allow you to control exactly when the load action (and thus latency) is invoked. The documentation for each resource explicitly lists its attributes.

Additionally, attributes may be reloaded after an action has been performed on the resource. For example, if the last_modified attribute of an S3 object is loaded and then a put action is called, then the next time you access last_modified it will reload the object’s metadata.


An action is a method which makes a call to the service. Actions may return a low-level response, a new resource instance or a list of new resource instances. Actions automatically set the resource identifiers as parameters, but allow you to pass additional parameters via keyword arguments. Examples of actions:

# SQS Queue
messages = queue.receive_messages()

# SQS Message
for message in messages:

# S3 Object
obj = s3.Object(bucket_name='boto3', key='')
response = obj.get()
data = response['Body'].read()

Examples of sending additional parameters:

# SQS Service
queue = sqs.get_queue_by_name(QueueName='test')

# SQS Queue


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


A reference is an attribute which may be None or a related resource instance. The resource instance does not share identifiers with its reference resource, that is, it is not a strict parent to child relationship. In relational terms, these can be considered many-to-one or one-to-one. Examples of references:

# EC2 Instance

In the above example, an EC2 instance may have exactly one associated subnet, and may have exactly one associated VPC. The subnet does not require the instance ID to exist, hence it is not a parent to child relationship.


A sub-resource is similar to a reference, but is a related class rather than an instance. Sub-resources, when instantiated, share identifiers with their parent. It is a strict parent-child relationship. In relational terms, these can be considered one-to-many. Examples of sub-resources:

queue = sqs.Queue(url='...')
message = queue.Message(receipt_handle='...')
print(queue.url == message.queue_url)

# S3
obj = bucket.Object(key='new_file.txt')

Because an SQS message cannot exist without a queue, and an S3 object cannot exist without a bucket, these are parent to child relationships.


A waiter is similar to an action. A waiter will poll the status of a resource and suspend execution until the resource reaches the state that is being polled for or a failure occurs while polling. Waiters automatically set the resource identifiers as parameters, but allow you to pass additional parameters via keyword arguments. Examples of waiters include:

# S3: Wait for a bucket to exist.

# EC2: Wait for an instance to reach the running state.

Multithreading or multiprocessing with resources#

Resource instances are not thread safe and should not be shared across threads or processes. These special classes contain additional meta data that cannot be shared. It’s recommended to create a new Resource for each thread or process:

import boto3
import boto3.session
import threading

class MyTask(threading.Thread):
    def run(self):
        # Here we create a new session per thread
        session = boto3.session.Session()

        # Next, we create a resource client using our thread's session object
        s3 = session.resource('s3')

        # Put your thread-safe code here

In the example above, each thread would have its own Boto3 session and its own instance of the S3 resource. This is a good idea because resources contain shared data when loaded and calling actions, accessing properties, or manually loading or reloading the resource can modify this data.