Azure Container App with EventHubConsumerClient gets stuck after idle period – requires restart to resume

Hi team,

I’ve deployed a Python app in Azure Container Apps which uses the EventHubConsumerClient from azure-eventhub to process events. The setup includes:

Azure Event Hub (with blob checkpointing via BlobCheckpointStore)

Azure Container App hosting the Python dispatcher

azure-eventhub async SDK (latest version)

Connection and error logging enabled (uamqp, azure.eventhub log levels: DEBUG)

Flask app with Celery for webhook delivery

Problem:
After being idle for a while (e.g., no events for ~10–30 minutes), the container stops receiving new events from Event Hub. No exceptions or disconnections are logged — it just silently stops processing. Only when I restart the container app, it starts consuming events again.

What I’ve tried:
Using client.receive(...) with starting_position='@latest'

Implementing a heartbeat logger to confirm the app is alive

Verifying no memory/CPU scaling issue in container logs

Logging Azure SDK + AMQP (uamqp) events for connection issues (none so far)

Confirmed Event Hub has new incoming events (checked via metrics)

Questions:
Is there a known connection idle timeout behavior in EventHub SDK for Python that silently drops without logging?

Should I be manually reinitializing the EventHubConsumerClient in a loop to handle idle disconnects?

Any known limitations or recommendations for long-lived consumers in Azure Container Apps with Event Hub?

Can Azure Monitor / Diagnostic settings be used to detect dropped or disconnected AMQP sessions from the Event Hub side?

Any insights, best practices, or SDK configurations to handle this properly would be greatly appreciated.

Thanks!

from azure.eventhub.aio import EventHubConsumerClient
from azure.eventhub.extensions.checkpointstoreblobaio import BlobCheckpointStore
import asyncio

async def on_event(partition_context, event):
    print(f"Received event from partition: {partition_context.partition_id}")
    await partition_context.update_checkpoint(event)

def run_dispatcher():
    checkpoint_store = BlobCheckpointStore.from_connection_string(
        "<blob-connection-string>", "<container-name>"
    )
    client = EventHubConsumerClient.from_connection_string(
        conn_str="<event-hub-conn-str>",
        consumer_group="$Default",
        eventhub_name="<event-hub-name>",
        checkpoint_store=checkpoint_store
    )

    loop = asyncio.get_event_loop()
    try:
        loop.run_until_complete(
            client.receive(
                on_event=on_event,
                starting_position="@latest"
            )
        )
    except Exception as e:
        print(f"Error in dispatcher: {e}")
    finally:
        loop.run_until_complete(client.close())
        loop.close()

Beginnen Sie mit dem bereitgestellten run_dispatcher-Einstiegspunkt und EventHubConsumerClient.receive und reproduzieren Sie dann den Fehler während einer Leerlaufphase mit dem beschriebenen Setup aus async client und BlobCheckpointStore. Überprüfen Sie den Consumer-Lebenszyklus und die Diagnosefunktionen des SDK rund um eine 10–30-minütige Leerlaufphase. Erledigt ist die Aufgabe, wenn der Fehler reproduziert und seine Ursache oder eine konkrete, getestete Lösung festgestellt wurde; das Issue nennt keine Repository-Dateien oder Tests.