Event Hub Trigger and Output Binding¶

Event Hubs is the streaming trigger for high-throughput append-only events. It is optimized for partitioned ingestion and consumer groups, making it the right fit for telemetry, clickstreams, and device/event pipelines rather than queue-style command processing.

Trigger¶

Use @app.event_hub_message_trigger(...) to consume events from a hub.

Key parameters: event_hub_name, connection, consumer_group, cardinality, and arg_name.

import azure.functions as func

app = func.FunctionApp()

@app.event_hub_message_trigger(
    arg_name="events",
    event_hub_name="telemetry",
    connection="EventHubConnection",
    consumer_group="$Default",
    cardinality="many",
)
def event_hub_example(events: list[func.EventHubEvent]) -> None:
    for event in events:
        print(event.get_body().decode("utf-8"))

Input Binding¶

Stream processors often enrich batches with point lookups before writing curated outputs.

import azure.functions as func

app = func.FunctionApp()

@app.event_hub_message_trigger(arg_name="events", event_hub_name="telemetry", connection="EventHubConnection")
@app.blob_input(arg_name="schema", path="schemas/telemetry.json", connection="AzureWebJobsStorage")
def enrich_stream(events: list[func.EventHubEvent], schema: bytes) -> None:
    print(f"schema_bytes={len(schema)} event_count={len(events)}")

Output Binding¶

Use the output binding to emit derived events to another hub.

import json
import azure.functions as func

app = func.FunctionApp()

@app.event_hub_message_trigger(arg_name="events", event_hub_name="telemetry", connection="EventHubConnection")
@app.event_hub_output(arg_name="out_events", event_hub_name="telemetry-curated", connection="EventHubConnection")
def republish_stream(events: list[func.EventHubEvent], out_events: func.Out[list[str]]) -> None:
    payloads = [json.dumps({"body": e.get_body().decode("utf-8")}) for e in events]
    out_events.set(payloads)

Configuration¶

Event Hub scaling and checkpoint behavior is host-controlled:

{
  "version": "2.0",
  "extensions": {
    "eventHubs": {
      "batchCheckpointFrequency": 1,
      "eventProcessorOptions": {
        "maxBatchSize": 100,
        "prefetchCount": 300,
        "loadBalancingUpdateInterval": "00:00:10"
      }
    }
  }
}

local.settings.json normally includes EventHubConnection plus AzureWebJobsStorage for leases/checkpoints when required by the extension version and hosting setup.

Scaling Behavior¶

Event Hub triggers scale by partition ownership. A partition is processed by only one active consumer in a consumer group at a time, so total parallelism is bounded by partition count. Ordering is preserved within a partition, and checkpoints determine how far the processor will replay after restarts or failures.

Common Pitfalls¶

Consumer-group changes create a new checkpoint lineage, which can look like duplicate processing.
cardinality="many" improves throughput but requires batch-safe code and memory budgeting.
Ordering across partitions does not exist; aggregate downstream by partition key if you need localized order.

https://learn.microsoft.com/azure/azure-functions/functions-bindings-event-hubs-trigger?pivots=programming-language-python&tabs=python-v2
https://learn.microsoft.com/azure/azure-functions/functions-bindings-event-hubs-output?pivots=programming-language-python&tabs=python-v2