azure-eventhub-py

$npx mdskill add microsoft/skills/azure-eventhub-py

Ingest and process high-throughput event streams using Azure Event Hubs.

  • Enables high-throughput event ingestion and streaming for big data pipelines.
  • Depends on azure-eventhub, azure-identity, and optional checkpoint store packages.
  • Uses DefaultAzureCredential for authentication and environment variables for configuration.
  • Provides producer and consumer clients with checkpointing for reliable event processing.

SKILL.md

.github/skills/azure-eventhub-pyView on GitHub ↗
---
name: azure-eventhub-py
description: |
  Azure Event Hubs SDK for Python streaming. Use for high-throughput event ingestion, producers, consumers, and checkpointing.
  Triggers: "event hubs", "EventHubProducerClient", "EventHubConsumerClient", "streaming", "partitions".
license: MIT
metadata:
  author: Microsoft
  version: "1.0.0"
  package: azure-eventhub
---

# Azure Event Hubs SDK for Python

Big data streaming platform for high-throughput event ingestion.

## Installation

```bash
pip install azure-eventhub azure-identity
# For checkpointing with blob storage
pip install azure-eventhub-checkpointstoreblob-aio
```

## Environment Variables

```bash
EVENT_HUB_FULLY_QUALIFIED_NAMESPACE=<namespace>.servicebus.windows.net  # Required for all auth methods
EVENT_HUB_NAME=my-eventhub  # Required for all auth methods
STORAGE_ACCOUNT_URL=https://<account>.blob.core.windows.net  # Required for checkpoint storage
CHECKPOINT_CONTAINER=checkpoints  # Required for checkpoint storage
AZURE_TOKEN_CREDENTIALS=prod # Required only if DefaultAzureCredential is used in production
```

## Authentication & Lifecycle

> **🔑 Two rules apply to every code sample below:**
>
> 1. **Prefer `DefaultAzureCredential`.** It works locally (Azure CLI / VS Code / Developer CLI) and in Azure (managed identity, workload identity) with no code change. Avoid connection strings, account/API keys — they bypass Entra audit and rotation.
>    - Local dev: `DefaultAzureCredential` works as-is.
>    - Production: set `AZURE_TOKEN_CREDENTIALS=prod` (or `AZURE_TOKEN_CREDENTIALS=<specific_credential>`) to constrain the credential chain to production-safe credentials.
> 2. **Wrap every client in a context manager** so HTTP transports, sockets, and token caches are released deterministically:
>    - Sync: `with <Client>(...) as client:`
>    - Async: `async with <Client>(...) as client:` **and** `async with DefaultAzureCredential() as credential:` (from `azure.identity.aio`)
>
> Snippets may abbreviate this setup, but production code should always follow both rules.

```python
from azure.identity import DefaultAzureCredential, ManagedIdentityCredential
from azure.eventhub import EventHubProducerClient, EventHubConsumerClient

# Local dev: DefaultAzureCredential. Production: set AZURE_TOKEN_CREDENTIALS=prod or AZURE_TOKEN_CREDENTIALS=<specific_credential>
credential = DefaultAzureCredential(require_envvar=True)
# Or use a specific credential directly in production:
# See https://learn.microsoft.com/python/api/overview/azure/identity-readme?view=azure-python#credential-classes
# credential = ManagedIdentityCredential()
namespace = "<namespace>.servicebus.windows.net"
eventhub_name = "my-eventhub"

# Producer
with EventHubProducerClient(
    fully_qualified_namespace=namespace,
    eventhub_name=eventhub_name,
    credential=credential
) as producer:
    # Use producer here (see following sections for operations)
    ...

# Consumer
with EventHubConsumerClient(
    fully_qualified_namespace=namespace,
    eventhub_name=eventhub_name,
    consumer_group="$Default",
    credential=credential
) as consumer:
    # Use consumer here (see following sections for operations)
    ...
```

## Client Types

| Client | Purpose |
|--------|---------|
| `EventHubProducerClient` | Send events to Event Hub |
| `EventHubConsumerClient` | Receive events from Event Hub |
| `BlobCheckpointStore` | Track consumer progress |

## Send Events

```python
from azure.eventhub import EventHubProducerClient, EventData
from azure.identity import DefaultAzureCredential

with EventHubProducerClient(
    fully_qualified_namespace="<namespace>.servicebus.windows.net",
    eventhub_name="my-eventhub",
    credential=DefaultAzureCredential()
) as producer:
    # Create batch (handles size limits)
    event_data_batch = producer.create_batch()
    
    for i in range(10):
        try:
            event_data_batch.add(EventData(f"Event {i}"))
        except ValueError:
            # Batch is full, send and create new one
            producer.send_batch(event_data_batch)
            event_data_batch = producer.create_batch()
            event_data_batch.add(EventData(f"Event {i}"))
    
    # Send remaining
    producer.send_batch(event_data_batch)
```

### Send to Specific Partition

```python
# By partition ID
event_data_batch = producer.create_batch(partition_id="0")

# By partition key (consistent hashing)
event_data_batch = producer.create_batch(partition_key="user-123")
```

## Receive Events

### Simple Receive

```python
from azure.eventhub import EventHubConsumerClient

def on_event(partition_context, event):
    print(f"Partition: {partition_context.partition_id}")
    print(f"Data: {event.body_as_str()}")
    partition_context.update_checkpoint(event)

with EventHubConsumerClient(
    fully_qualified_namespace="<namespace>.servicebus.windows.net",
    eventhub_name="my-eventhub",
    consumer_group="$Default",
    credential=DefaultAzureCredential()
) as consumer:
    consumer.receive(
        on_event=on_event,
        starting_position="-1",  # Beginning of stream
    )
```

### With Blob Checkpoint Store (Production)

```python
from azure.eventhub import EventHubConsumerClient
from azure.eventhub.extensions.checkpointstoreblob import BlobCheckpointStore
from azure.identity import DefaultAzureCredential

checkpoint_store = BlobCheckpointStore(
    blob_account_url="https://<account>.blob.core.windows.net",
    container_name="checkpoints",
    credential=DefaultAzureCredential()
)

with EventHubConsumerClient(
    fully_qualified_namespace="<namespace>.servicebus.windows.net",
    eventhub_name="my-eventhub",
    consumer_group="$Default",
    credential=DefaultAzureCredential(),
    checkpoint_store=checkpoint_store
) as consumer:
    def on_event(partition_context, event):
        print(f"Received: {event.body_as_str()}")
        # Checkpoint after processing
        partition_context.update_checkpoint(event)

    consumer.receive(on_event=on_event)
```

## Async Client

```python
from azure.eventhub.aio import EventHubProducerClient, EventHubConsumerClient
from azure.identity.aio import DefaultAzureCredential
import asyncio

async def send_events():
    credential = DefaultAzureCredential()
    
    async with EventHubProducerClient(
        fully_qualified_namespace="<namespace>.servicebus.windows.net",
        eventhub_name="my-eventhub",
        credential=credential
    ) as producer:
        batch = await producer.create_batch()
        batch.add(EventData("Async event"))
        await producer.send_batch(batch)

async def receive_events():
    async def on_event(partition_context, event):
        print(event.body_as_str())
        await partition_context.update_checkpoint(event)
    
    async with EventHubConsumerClient(
        fully_qualified_namespace="<namespace>.servicebus.windows.net",
        eventhub_name="my-eventhub",
        consumer_group="$Default",
        credential=DefaultAzureCredential()
    ) as consumer:
        await consumer.receive(on_event=on_event)

asyncio.run(send_events())
```

## Event Properties

```python
event = EventData("My event body")

# Set properties
event.properties = {"custom_property": "value"}
event.content_type = "application/json"

# Read properties (on receive)
print(event.body_as_str())
print(event.sequence_number)
print(event.offset)
print(event.enqueued_time)
print(event.partition_key)
```

## Get Event Hub Info

```python
with producer:
    info = producer.get_eventhub_properties()
    print(f"Name: {info['name']}")
    print(f"Partitions: {info['partition_ids']}")
    
    for partition_id in info['partition_ids']:
        partition_info = producer.get_partition_properties(partition_id)
        print(f"Partition {partition_id}: {partition_info['last_enqueued_sequence_number']}")
```

## Best Practices

1. **Pick sync OR async and stay consistent.** Do not mix `azure.xxx` sync clients with `azure.xxx.aio` async clients in the same call path. Choose one mode per module.
2. **Always use context managers for clients and async credentials.** Wrap every client in `with Client(...) as client:` (sync) or `async with Client(...) as client:` (async) for proper cleanup. For async `DefaultAzureCredential` from `azure.identity.aio`, also use `async with credential:` so tokens and transports are cleaned up.
3. **Use `DefaultAzureCredential`** for portable auth across local dev and Azure (avoid connection strings / API keys when possible).
4. **Use batches** for sending multiple events
5. **Use checkpoint store** in production for reliable processing
6. **Use async client** for high-throughput scenarios
7. **Use partition keys** for ordered delivery within a partition
8. **Handle batch size limits** — catch ValueError when batch is full
9. **Set appropriate consumer groups** for different applications

## Reference Files

| File | Contents |
|------|----------|
| [references/checkpointing.md](references/checkpointing.md) | Checkpoint store patterns, blob checkpointing, checkpoint strategies |
| [references/partitions.md](references/partitions.md) | Partition management, load balancing, starting positions |
| [scripts/setup_consumer.py](scripts/setup_consumer.py) | CLI for Event Hub info, consumer setup, and event sending/receiving |

More from microsoft/skills

SkillDescription
agent-framework-azure-ai-pyBuild Azure AI Foundry agents using the Microsoft Agent Framework Python SDK (agent-framework-azure-ai). Use when creating persistent agents with AzureAIAgentsProvider, using hosted tools (code interpreter, file search, web search), integrating MCP servers, managing conversation threads, or implementing streaming responses. Covers function tools, structured outputs, and multi-tool agents.
airunway-aks-setupSet up AI Runway on AKS — from bare cluster to running model. Covers cluster verification, controller install, GPU assessment, provider setup, and first deployment. WHEN: \"setup AI Runway\", \"onboard AKS cluster\", \"install AI Runway\", \"airunway setup\", \"deploy model to AKS\", \"GPU inference on AKS\", \"KAITO setup on AKS\", \"run LLM on AKS\", \"vLLM on AKS\", \"set up model serving on AKS\", \"AI Runway controller\".
appinsights-instrumentationGuidance for instrumenting webapps with Azure Application Insights. Provides telemetry patterns, SDK setup, and configuration references. WHEN: how to instrument app, App Insights SDK, telemetry patterns, what is App Insights, Application Insights guidance, instrumentation examples, APM best practices.
applicationinsights-web-tsInstrument browser/web apps with the Application Insights JavaScript SDK (@microsoft/applicationinsights-web). Use for Real User Monitoring (RUM) — page views, clicks, AJAX/fetch dependencies, exceptions, custom events, and browser-side GenAI agent traces correlated to backend OpenTelemetry traces. Covers SDK Loader Script and npm setup, framework extensions (React, React Native, Angular), Click Analytics, telemetry initializers, and OTel GenAI semantic conventions for agent/tool/model spans emitted from the browser.
azure-aiUse for Azure AI: Search, Speech, OpenAI, Document Intelligence. Helps with search, vector/hybrid search, speech-to-text, text-to-speech, transcription, OCR. WHEN: AI Search, query search, vector search, hybrid search, semantic search, speech-to-text, text-to-speech, transcribe, OCR, convert text to speech.
azure-ai-agents-persistent-dotnet|
azure-ai-agents-persistent-java|
azure-ai-anomalydetector-javaBuild anomaly detection applications with Azure AI Anomaly Detector SDK for Java. Use when implementing univariate/multivariate anomaly detection, time-series analysis, or AI-powered monitoring.
azure-ai-contentsafety-javaBuild content moderation applications with Azure AI Content Safety SDK for Java. Use when implementing text/image analysis, blocklist management, or harm detection for hate, violence, sexual content, and self-harm.
azure-ai-contentsafety-py|