AWS SQS Listener

This package takes care of the boilerplate involved in listening to an SQS queue, as well as sending messages to a queue. Works with python 2.7 & 3.6+.

Installation

pip install pySqsListener

Listening to a queue

Using the listener is very straightforward - just inherit from the SqsListener class and implement the handle_message() method. The queue will be created at runtime if it doesn't already exist. You can also specify an error queue to automatically push any errors to.

Here is a basic code sample:

Standard Listener

from sqs_listener import SqsListener

class MyListener(SqsListener):
    def handle_message(self, body, attributes, messages_attributes):
        run_my_function(body['param1'], body['param2'])

listener = MyListener('my-message-queue', error_queue='my-error-queue', region_name='us-east-1')
listener.listen()

Error Listener

from sqs_listener import SqsListener
class MyErrorListener(SqsListener):
    def handle_message(self, body, attributes, messages_attributes):
        save_to_log(body['exception_type'], body['error_message']

error_listener = MyErrorListener('my-error-queue')
error_listener.listen()

The options available as kwargs are as follows:

error_queue (str) - name of queue to push errors.
force_delete (boolean) - delete the message received from the queue, whether or not the handler function is successful. By default the message is deleted only if the handler function returns with no exceptions
interval (int) - number of seconds in between polls. Set to 60 by default
visibility_timeout (str) - Number of seconds the message will be invisible ('in flight') after being read. After this time interval it reappear in the queue if it wasn't deleted in the meantime. Set to '600' (10 minutes) by default
error_visibility_timeout (str) - Same as previous argument, for the error queue. Applicable only if the error_queue argument is set, and the queue doesn't already exist.
wait_time (int) - number of seconds to wait for a message to arrive (for long polling). Set to 0 by default to provide short polling.
max_number_of_messages (int) - Max number of messages to receive from the queue. Set to 1 by default, max is 10
message_attribute_names (list) - message attributes by which to filter messages
attribute_names (list) - attributes by which to filter messages (see boto docs for difference between these two)
region_name (str) - AWS region name (defaults to us-east-1)
queue_url (str) - overrides queue parameter. Mostly useful for getting around this bug in the boto library

Running as a Daemon

Typically, in a production environment, you'll want to listen to an SQS queue with a daemonized process. The simplest way to do this is by running the listener in a detached process. On a typical Linux distribution it might look like this:

nohup python my_listener.py > listener.log &

And saving the resulting process id for later (for stopping the listener via the kill command).

A more complete implementation can be achieved easily by inheriting from the package's Daemon class and overriding the run() method.

The sample_daemon.py file in the source root folder provides a clear example for achieving this. Using this example, you can run the listener as a daemon with the command python sample_daemon.py start. Similarly, the command python sample_daemon.py stop will stop the process. You'll most likely need to run the start script using sudo.

Logging

The listener and launcher instances push all their messages to a logger instance, called 'sqs_listener'. In order to view the messages, the logger needs to be redirected to stdout or to a log file.

For instance:

logger = logging.getLogger('sqs_listener')
logger.setLevel(logging.INFO)

sh = logging.StreamHandler(sys.stdout)
sh.setLevel(logging.INFO)

formatstr = '[%(asctime)s - %(name)s - %(levelname)s]  %(message)s'
formatter = logging.Formatter(formatstr)

sh.setFormatter(formatter)
logger.addHandler(sh)

Or to a log file:

logger = logging.getLogger('sqs_listener')
logger.setLevel(logging.INFO)

sh = logging.FileHandler('mylog.log')
sh.setLevel(logging.INFO)

formatstr = '[%(asctime)s - %(name)s - %(levelname)s]  %(message)s'
formatter = logging.Formatter(formatstr)

sh.setFormatter(formatter)
logger.addHandler(sh)

Sending messages

In order to send a message, instantiate an SqsLauncher with the name of the queue. By default an exception will be raised if the queue doesn't exist, but it can be created automatically if the create_queue parameter is set to true. In such a case, there's also an option to set the newly created queue's VisibilityTimeout via the third parameter.

After instantiation, use the launch_message() method to send the message. The message body should be a dict, and additional kwargs can be specified as stated in the SQS docs. The method returns the response from SQS.

Launcher Example

from sqs_launcher import SqsLauncher

launcher = SqsLauncher('my-queue')
response = launcher.launch_message({'param1': 'hello', 'param2': 'world'})

Important Notes

The environment variable AWS_ACCOUNT_ID must be set, in addition to the environment having valid AWS credentials (via environment variables or a credentials file) or if running in an aws ec2 instance a role attached with the required permissions.
For both the main queue and the error queue, if the queue doesn’t exist (in the specified region), it will be created at runtime.
The error queue receives only two values in the message body: exception_type and error_message. Both are of type str
If the function that the listener executes involves connecting to a database, you should explicitly close the connection at the end of the function. Otherwise, you're likely to get an error like this: OperationalError(2006, 'MySQL server has gone away')
Either the queue name or the queue url should be provided. When both are provided the queue url is used and the queue name is ignored.

Contributing

Fork the repo and make a pull request.

la-mar/python-sqs-listener