Azure Event Hubs client library for Python

Azure Event Hubs is a highly scalable publish-subscribe service that can ingest millions of events per second and stream them to multiple consumers. This lets you process and analyze the massive amounts of data produced by your connected devices and applications. Once Event Hubs has collected the data, you can retrieve, transform, and store it by using any real-time analytics provider or with batching/storage adapters. If you would like to know more about Azure Event Hubs, you may wish to review: What is Event Hubs?

The Azure Event Hubs client library allows for publishing and consuming of Azure Event Hubs events and may be used to:

• Emit telemetry about your application for business intelligence and diagnostic purposes.
• Publish facts about the state of your application which interested parties may observe and use as a trigger for taking action.
• Observe interesting operations and interactions happening within your business or other ecosystem, allowing loosely coupled systems to interact without the need to bind them together.
• Receive events from one or more publishers, transform them to better meet the needs of your ecosystem, then publish the transformed events to a new stream for consumers to observe.

Source code | Package (PyPi) | API reference documentation | Product documentation

Getting started

Install the package

Install the Azure Event Hubs client library for Python with pip:

$ pip install --pre azure-eventhub

Prerequisites

• Python 2.7, 3.5 or later.

• Microsoft Azure Subscription: To use Azure services, including Azure Event Hubs, you'll need a subscription. If you do not have an existing Azure account, you may sign up for a free trial or use your MSDN subscriber benefits when you create an account.

• Event Hubs namespace with an Event Hub: To interact with Azure Event Hubs, you'll also need to have a namespace and Event Hub available. If you are not familiar with creating Azure resources, you may wish to follow the step-by-step guide for creating an Event Hub using the Azure portal. There, you can also find detailed instructions for using the Azure CLI, Azure PowerShell, or Azure Resource Manager (ARM) templates to create an Event Hub.

Authenticate the client

Interaction with Event Hubs starts with an instance of the EventHubClient class. You need the host name, SAS/AAD credential and event hub name to instantiate the client object.

Obtain a connection string

For the Event Hubs client library to interact with an Event Hub, it will need to understand how to connect and authorize with it. The easiest means for doing so is to use a connection string, which is created automatically when creating an Event Hubs namespace. If you aren't familiar with shared access policies in Azure, you may wish to follow the step-by-step guide to get an Event Hubs connection string.

Create client

There are several ways to instantiate the EventHubClient object and the following code snippets demonstrate two ways:

Create client from connection string:

from azure.eventhub import EventHubClient

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient.from_connection_string(connection_str, event_hub_path)

• The from_connection_string method takes the connection string of the form Endpoint=sb://<yournamespace>.servicebus.windows.net/;SharedAccessKeyName=<yoursharedaccesskeyname>;SharedAccessKey=<yoursharedaccesskey> and entity name to your Event Hub instance. You can get the connection string from the Azure portal.

Create client using the azure-identity library:

from azure.eventhub import EventHubClient
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()

host = '<< HOSTNAME OF THE EVENT HUB >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient(host, event_hub_path, credential)

• This constructor takes the host name and entity name of your Event Hub instance and credential that implements the TokenCredential interface. There are implementations of the TokenCredential interface available in the azure-identity package. The host name is of the format <yournamespace.servicebus.windows.net>.

Key concepts

• An Event Hub client is the primary interface for developers interacting with the Event Hubs client library, allowing for inspection of Event Hub metadata and providing a guided experience towards specific Event Hub operations such as the creation of producers and consumers.

• An Event Hub producer is a source of telemetry data, diagnostics information, usage logs, or other log data, as part of an embedded device solution, a mobile device application, a game title running on a console or other device, some client or server based business solution, or a web site.

• An Event Hub consumer picks up such information from the Event Hub and processes it. Processing may involve aggregation, complex computation, and filtering. Processing may also involve distribution or storage of the information in a raw or transformed fashion. Event Hub consumers are often robust and high-scale platform infrastructure parts with built-in analytics capabilities, like Azure Stream Analytics, Apache Spark, or Apache Storm.

• A partition is an ordered sequence of events that is held in an Event Hub. Azure Event Hubs provides message streaming through a partitioned consumer pattern in which each consumer only reads a specific subset, or partition, of the message stream. As newer events arrive, they are added to the end of this sequence. The number of partitions is specified at the time anEvent Hub is created and cannot be changed.

• A consumer group is a view of an entire Event Hub. Consumer groups enable multiple consuming applications to each have a separate view of the event stream, and to read the stream independently at their own pace and from their own position. There can be at most 5 concurrent readers on a partition per consumer group; however it is recommended that there is only one active consumer for a given partition and consumer group pairing. Each active reader receives all of the events from its partition; if there are multiple readers on the same partition, then they will receive duplicate events.

For more concepts and deeper discussion, see: Event Hubs Features. Also, the concepts for AMQP are well documented in OASIS Advanced Messaging Queuing Protocol (AMQP) Version 1.0.

Examples

The following sections provide several code snippets covering some of the most common Event Hubs tasks, including:

• Inspect an Event Hub
• Publish events to an Event Hub
• Consume events from an Event Hub
• Async publish events to an Event Hub
• Async consume events from an Event Hub
• Consume events using an Event Processor

Inspect an Event Hub

Get the partition ids of an Event Hub.

from azure.eventhub import EventHubClient

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient.from_connection_string(connection_str, event_hub_path)
partition_ids = client.get_partition_ids()

Publish events to an Event Hub

Publish events to an Event Hub.

Send a single event or an array of events

from azure.eventhub import EventHubClient, EventData

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient.from_connection_string(connection_str, event_hub_path)
producer = client.create_producer(partition_id="0")

try:
 	event_list = []
 	for i in range(10):
 		event_list.append(EventData(b"A single event"))

 	with producer:
 	    producer.send(event_list)
except:
	raise
finally:
    pass

Send a batch of events

Use the create_batch method on EventHubProcuer to create an EventDataBatch object which can then be sent using the send method. Events may be added to the EventDataBatch using the try_add method until the maximum batch size limit in bytes has been reached.

from azure.eventhub import EventHubClient, EventData

try:
    connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
    event_hub_path = '<< NAME OF THE EVENT HUB >>'
    client = EventHubClient.from_connection_string(connection_str, event_hub_path)
    producer = client.create_producer(partition_id="0")

    event_data_batch = producer.create_batch(max_size=10000)
    can_add = True
    while can_add:
        try:
            event_data_batch.try_add(EventData('Message inside EventBatchData'))
        except ValueError:
            can_add = False  # EventDataBatch object reaches max_size.

    with producer:
        producer.send(event_data_batch)
except:
	raise
finally:
    pass

Consume events from an Event Hub

Consume events from an Event Hub.

import logging
from azure.eventhub import EventHubClient, EventData, EventPosition

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient.from_connection_string(connection_str, event_hub_path)
consumer = client.create_consumer(consumer_group="$default", partition_id="0", event_position=EventPosition("-1"))

try:
    logger = logging.getLogger("azure.eventhub")
    with consumer:
        received = consumer.receive(max_batch_size=100, timeout=5)
        for event_data in received:
            logger.info("Message received:{}".format(event_data))
except:
    raise
finally:
    pass

Async publish events to an Event Hub

Publish events to an Event Hub asynchronously.

Send a single event or an array of events

from azure.eventhub.aio import EventHubClient
from azure.eventhub import EventData

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient.from_connection_string(connection_str, event_hub_path)
producer = client.create_producer(partition_id="0")

try:
 	event_list = []
 	for i in range(10):
 		event_list.append(EventData(b"A single event"))

	async with producer:
		await producer.send(event_list)  # Send a list of events
		await producer.send(EventData(b"A single event"))  # Send a single event
except:
	raise
finally:
    pass

Send a batch of events

Use the create_batch method on EventHubProcuer to create an EventDataBatch object which can then be sent using the send method. Events may be added to the EventDataBatch using the try_add method until the maximum batch size limit in bytes has been reached.

from azure.eventhub.aio import EventHubClient
from azure.eventhub import EventData

try:
    connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
    event_hub_path = '<< NAME OF THE EVENT HUB >>'
    client = EventHubClient.from_connection_string(connection_str, event_hub_path)
    producer = client.create_producer(partition_id="0")

    event_data_batch = await producer.create_batch(max_size=10000)
    can_add = True
    while can_add:
        try:
            event_data_batch.try_add(EventData('Message inside EventBatchData'))
        except ValueError:
            can_add = False  # EventDataBatch object reaches max_size.

    async with producer:
        await producer.send(event_data_batch)
except:
	raise
finally:
    pass

Async consume events from an Event Hub

Consume events asynchronously from an EventHub.

import logging
from azure.eventhub.aio import EventHubClient
from azure.eventhub import EventData, EventPosition

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'
event_hub_path = '<< NAME OF THE EVENT HUB >>'
client = EventHubClient.from_connection_string(connection_str, event_hub_path)
consumer = client.create_consumer(consumer_group="$default", partition_id="0", event_position=EventPosition("-1"))

try:
    logger = logging.getLogger("azure.eventhub")
    async with consumer:
        received = await consumer.receive(max_batch_size=100, timeout=5)
        for event_data in received:
            logger.info("Message received:{}".format(event_data))
except:
    raise
finally:
    pass

Consume events using an Event Processor

EventProcessor is a high level construct which internally uses the EventHubConsumer mentioned in previous examples to receive events from multiple partitions at once.

Typically, Event Processor based applications consist of one or more instances of EventProcessor instances which have been configured to consume events from the same Event Hub and consumer group. They balance the workload across different instances by distributing the partitions to be processed among themselves. They also allow the user to track progress when events are processed using checkpoints.

The EventProcessor will delegate the processing of events to a PartitionProcessor that you provide, allowing you to focus on business logic while the EventProcessor holds responsibility for managing the underlying consumer operations including checkpointing and load balancing.

A checkpoint is meant to represent the last successfully processed event by the user from a particular partition of a consumer group in an Event Hub instance.The EventProcessor uses an instance of PartitionManager to update checkpoints and to store the relevant information required by the load balancing algorithm. While for the purposes of getting started you can use the SamplePartitionManager that is shipped out of the box from this library, it is recommended to use a persistent store when running in production. Search pypi with the prefix azure-eventhub-checkpointstore to find packages that support this and use the PartitionManager implementation from one such package.

In the below example, we create an instance of EventProcessor and use an SamplePartitionManager.

Azure Blob Storage Partition Manager is one of the PartitionManager implementation we provide that applies Azure Blob Storage as the persistent store.

import asyncio

from azure.eventhub.aio import EventHubClient
from azure.eventhub.aio.eventprocessor import EventProcessor, PartitionProcessor, SamplePartitionManager

connection_str = '<< CONNECTION STRING FOR THE EVENT HUBS NAMESPACE >>'

async def do_operation(event):
    # do some sync or async operations. If the operation is i/o intensive, async will have better performance
    print(event)

class MyPartitionProcessor(PartitionProcessor):
    async def process_events(self, events, partition_context):
        if events:
            await asyncio.gather(*[do_operation(event) for event in events])
            await partition_context.update_checkpoint(events[-1].offset, events[-1].sequence_number)

async def main():
    client = EventHubClient.from_connection_string(connection_str, receive_timeout=5, retry_total=3)
    partition_manager = SamplePartitionManager()  # in-memory or file based PartitionManager
    try:
        event_processor = EventProcessor(client, "$default", MyPartitionProcessor, partition_manager)
        asyncio.ensure_future(event_processor.start())
        await asyncio.sleep(60)
        await event_processor.stop()
    finally:
        await partition_manager.close()

if __name__ == '__main__':
    loop = asyncio.get_event_loop()
    loop.run_until_complete(main())

Troubleshooting

General

The Event Hubs APIs generate the following exceptions.

• AuthenticationError: Failed to authenticate because of wrong address, SAS policy/key pair, SAS token or azure identity.
• ConnectError: Failed to connect to the EventHubs. The AuthenticationError is a type of ConnectError.
• ConnectionLostError: Lose connection after a connection has been built.
• EventDataError: The EventData to be sent fails data validation. For instance, this error is raised if you try to send an EventData that is already sent.
• EventDataSendError: The Eventhubs service responds with an error when an EventData is sent.
• OperationTimeoutError: EventHubConsumer.send() times out.
• EventHubError: All other Eventhubs related errors. It is also the root error class of all the above mentioned errors.

Next steps

Examples

These are the samples in our repo demonstraing the usage of the library.

• ./examples/send.py - use producer to publish events
• ./examples/recv.py - use consumer to consume events
• ./examples/async_examples/send_async.py - async/await support of a producer
• ./examples/async_examples/recv_async.py - async/await support of a consumer
• ./examples/eventprocessor/event_processor_example.py - event processor

Documentation

Reference documentation is available at https://azure.github.io/azure-sdk-for-python/ref/azure.eventhub.

Logging

• Enable azure.eventhub logger to collect traces from the library.
• Enable uamqp logger to collect traces from the underlying uAMQP library.
• Enable AMQP frame level trace by setting network_tracing=True when creating the client.

Provide Feedback

If you encounter any bugs or have suggestions, please file an issue in the Issues section of the project.

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Release History

5.0.0b3 (2019-09-10)

New features

• Added support for automatic load balancing among multiple EventProcessor.
• Added BlobPartitionManager which implements PartitionManager.
  • Azure Blob Storage is applied for storing data used by EventProcessor.
  • Packaged separately as a plug-in to EventProcessor.
  • For details, please refer to Azure Blob Storage Partition Manager.
• Added property system_properties on EventData.

Breaking changes

• Removed constructor method of PartitionProcessor. For initialization please implement the method initialize.
• Replaced CheckpointManager by PartitionContext.
  • PartitionContext has partition context information and method update_checkpoint.
• Updated all methods of PartitionProcessor to include PartitionContext as part of the arguments.
• Updated accessibility of class members in EventHub/EventHubConsumer/EventHubProducerto be private.
• Moved azure.eventhub.eventprocessor under aio package, which now becomes azure.eventhub.aio.eventprocessor.

5.0.0b2 (2019-08-06)

New features

• Added method create_batch on the EventHubProducer to create an EventDataBatch that can then be used to add events until the maximum size is reached.
  • This batch object can then be used in the send() method to send all the added events to Event Hubs.
  • This allows publishers to build batches without the possibility of encountering the error around the message size exceeding the supported limit when sending events.
  • It also allows publishers with bandwidth concerns to control the size of each batch published.
• Added new configuration parameters for exponential delay between retry operations.
  • retry_total: The total number of attempts to redo the failed operation.
  • backoff_factor: The delay time factor.
  • backoff_max: The maximum delay time in total.
• Added support for context manager on EventHubClient.
• Added new error type OperationTimeoutError for send operation.
• Introduced a new class EventProcessor which replaces the older concept of Event Processor Host. This early preview is intended to allow users to test the new design using a single instance of EventProcessor. The ability to checkpoints to a durable store will be added in future updates.
  • EventProcessor: EventProcessor creates and runs consumers for all partitions of the eventhub.
  • PartitionManager: PartitionManager defines the interface for getting/claiming ownerships of partitions and updating checkpoints.
  • PartitionProcessor: PartitionProcessor defines the interface for processing events.
  • CheckpointManager: CheckpointManager takes responsibility for updating checkpoints during events processing.

Breaking changes

• EventProcessorHost was replaced by EventProcessor, please read the new features for details.
• Replaced max_retries configuration parameter of the EventHubClient with retry_total.

5.0.0b1 (2019-06-25)

Version 5.0.0b1 is a preview of our efforts to create a client library that is user friendly and idiomatic to the Python ecosystem. The reasons for most of the changes in this update can be found in the Azure SDK Design Guidelines for Python. For more information, please visit https://aka.ms/azure-sdk-preview1-python.

New features

• Added new configuration parameters for creating EventHubClient.
  • credential: The credential object used for authentication which implements TokenCredential interface of getting tokens.
  • transport_type: The type of transport protocol that will be used for communicating with the Event Hubs service.
  • max_retries: The max number of attempts to redo the failed operation when an error happened.
  • for detailed information about the configuration parameters, please read the reference documentation.
• Added new methods get_partition_properties and get_partition_ids to EventHubClient.
• Added support for http proxy.
• Added support for authentication using azure-identity credential.
• Added support for transport using AMQP over WebSocket.

Breaking changes

• New error hierarchy
  • azure.error.EventHubError
  • azure.error.ConnectionLostError
  • azure.error.ConnectError
  • azure.error.AuthenticationError
  • azure.error.EventDataError
  • azure.error.EventDataSendError
• Renamed Sender/Receiver to EventHubProducer/EventHubConsumer.
  • Renamed add_sender to create_producer and add_receiver to create_consumer in EventHubClient.
  • EventHubConsumer is now iterable.
• Rename class azure.eventhub.Offset to azure.eventhub.EventPosition.
• Rename method get_eventhub_info to get_properties of EventHubClient.
• Reorganized connection management, EventHubClient is no longer responsible for opening/closing EventHubProducer/EventHubConsumer.
  • Each EventHubProducer/EventHubConsumer is responsible for its own connection management.
  • Added support for context manager on EventHubProducer and EventHubConsumer.
• Reorganized async APIs into "azure.eventhub.aio" namespace and rename to drop the "_async" suffix.
• Updated uAMQP dependency to 1.2.

1.3.1 (2019-02-28)

BugFixes

• Fixed bug where datetime offset filter was using a local timestamp rather than UTC.
• Fixed stackoverflow error in continuous connection reconnect attempts.

1.3.0 (2019-01-29)

BugFixes

• Added support for auto reconnect on token expiration and other auth errors (issue #89).

Features

• Added ability to create ServiceBusClient from an existing SAS auth token, including providing a function to auto-renew that token on expiry.
• Added support for storing a custom EPH context value in checkpoint (PR #84, thanks @konstantinmiller)

1.2.0 (2018-11-29)

• Support for Python 2.7 in azure.eventhub module (azure.eventprocessorhost will not support Python 2.7).
• Parse EventData.enqueued_time as a UTC timestamp (issue #72, thanks @vjrantal)

1.1.1 (2018-10-03)

• Fixed bug in Azure namespace package.

1.1.0 (2018-09-21)

• Changes to AzureStorageCheckpointLeaseManager parameters to support other connection options (issue #61):

  • The storage_account_name, storage_account_key and lease_container_name arguments are now optional keyword arguments.
  • Added a sas_token argument that must be specified with storage_account_name in place of storage_account_key.
  • Added an endpoint_suffix argument to support storage endpoints in National Clouds.
  • Added a connection_string argument that, if specified, overrides all other endpoint arguments.
  • The lease_container_name argument now defaults to "eph-leases" if not specified.

• Fix for clients failing to start if run called multipled times (issue #64).

• Added convenience methods body_as_str and body_as_json to EventData object for easier processing of message data.

1.0.0 (2018-08-22)

• API stable.
• Renamed internal _async module to async_ops for docs generation.
• Added optional auth_timeout parameter to EventHubClient and EventHubClientAsync to configure how long to allow for token negotiation to complete. Default is 60 seconds.
• Added optional send_timeout parameter to EventHubClient.add_sender and EventHubClientAsync.add_async_sender to determine the timeout for Events to be successfully sent. Default value is 60 seconds.
• Reformatted logging for performance.

0.2.0 (2018-08-06)

• Stability improvements for EPH.

• Updated uAMQP version.

• Added new configuration options for Sender and Receiver; keep_alive and auto_reconnect. These flags have been added to the following:

  • EventHubClient.add_receiver
  • EventHubClient.add_sender
  • EventHubClientAsync.add_async_receiver
  • EventHubClientAsync.add_async_sender
  • EPHOptions.keey_alive_interval
  • EPHOptions.auto_reconnect_on_error

0.2.0rc2 (2018-07-29)

• Breaking change EventData.offset will now return an object of type ~uamqp.common.Offset rather than str. The original string value can be retrieved from ~uamqp.common.Offset.value.
• Each sender/receiver will now run in its own independent connection.
• Updated uAMQP dependency to 0.2.0
• Fixed issue with IoTHub clients not being able to retrieve partition information.
• Added support for HTTP proxy settings to both EventHubClient and EPH.
• Added error handling policy to automatically reconnect on retryable error.
• Added keep-alive thread for maintaining an unused connection.

0.2.0rc1 (2018-07-06)

• Breaking change Restructured library to support Python 3.7. Submodule async has been renamed and all classes from this module can now be imported from azure.eventhub directly.
• Breaking change Removed optional callback argument from Receiver.receive and AsyncReceiver.receive.
• Breaking change EventData.properties has been renamed to EventData.application_properties. This removes the potential for messages to be processed via callback for not yet returned in the batch.
• Updated uAMQP dependency to v0.1.0
• Added support for constructing IoTHub connections.
• Fixed memory leak in receive operations.
• Dropped Python 2.7 wheel support.

0.2.0b2 (2018-05-29)

• Added namespace_suffix to EventHubConfig() to support national clouds.
• Added device_id attribute to EventData to support IoT Hub use cases.
• Added message header to workaround service bug for PartitionKey support.
• Updated uAMQP dependency to vRC1.

0.2.0b1 (2018-04-20)

• Updated uAMQP to latest version.
• Further testing and minor bug fixes.

0.2.0a2 (2018-04-02)

• Updated uAQMP dependency.

0.2.0a1 (unreleased)

• Swapped out Proton dependency for uAMQP.