Event Hub Trigger does not process the messages of a partition sequentially

Hi,

we are using the Event Hub trigger in a Java-based Azure function for processing messages from IoT Hub. We are relying on the fact that message are processed from a partition in the order that they were written to the Event Hub partition, or more specifically that messages with the same partition key are sequentially processed by a single thread. This property of Event Hub partitions is referenced in several places of the documentation, e.g.

> Within each partition, events remain in production order

> Use keys when consumers need to receive events in production order. Since all events with the same key go to the same partition, events with key values can maintain their order during processing"

Source: https://docs.microsoft.com/en-us/azure/architecture/reference-architectures/event-hubs/partitioning-in-event-hubs-and-kafka

> In some scenarios, the ordering of events can be important. For example, you may want your back-end system to process an update command before a delete command. In this scenario, a client application sends events to a specific partition so that the ordering is preserved. When a consumer application consumes these events from the partition, they are read in order

Source: https://docs.microsoft.com/en-us/azure/event-hubs/event-hubs-availability-and-consistency?tabs=dotnet

We have found that at least in Java Azure Functions using the Event Hub trigger this is not the case. Instead the messages seem to be dispatched to a thread pool, so that messages of the same partition are often executed concurrently and out of order.

Please advise if this behavior can be changed using configuration. Coming from Apache Kafka that uses the same partitioning and concurrency concept, we were very surprised that upholding the message and processing ordering guarantee per partition is not the default behavior.

Comience revisando el comportamiento del Java Azure Functions Event Hub trigger descrito en el informe, centrándose en la asignación de particiones, el despacho mediante el thread-pool y la configuración disponible. Confirme si los mensajes de una misma partición pueden ejecutarse de forma concurrente y fuera de orden, y documente después la configuración admitida o el comportamiento que debe cambiar.

Escrito por el modelo de indexación a partir del texto del issue.