A container that acts like an event hub so that in local/dev/test/ci environments it's easy to spin up and tear down application stacks that are dependent on event hub.

Currently, the dedicated cluster picks up the internal monitoring namespace traffic into the aggregated metrics. As this is the only place that the internal workings shows up, it is confusing to see thousands of messages ingress into a cluster with no user defined namespaces or hubs on it.

The internals should either be hidden and abstracted, or separated out from user metrics.

Currently, when a new cluster is created, it shows metrics from notably before its creation timestamp which is confusing.

New clusters should only show metrics related to workloads that occurred on them from their creation time forward.

Currently, if you deploy a dedicated cluster you have to wait four hours to delete it. If the four hour window is related to the minimum billing period, the cluster should still be able to be deleted (even though you will be billed for it).

If it's not tied to the billing window, it would be nice to be able to pend the delete until whatever technical systems are ready and then let the system take over, rather than letting it fail.

When activating the built-in capture feature in Event Hub, messages are stored in Blobs using AVRO format. In AVRO, system prop EnqueuedTimeUTC is serialized as a string, and the odd part is that milliseconds are not there. Basically, capture feature is truncating the original information, which it should not do

Event Hub kafka enabled does not allows KSQL client to be connect

Need to disable TLS 1.0, 1.1 and 3DES from Azure Event hubs, Service bus and relay.

Event hub namespaces have a hard limit of 10 hubs. It would be great to be able to raise that limit to 100 or 1000 or whatever via a support request.

We use Azure Event Hubs with Kafka. A namespace for us is a kafka server and a hub is a kafka topic. Now we have to split topics per kafka server based on Azure's 10 hubs hard limit, which is counterintuitive and complicates our setup.

Use case: Output from ARM template the primary connection string of an event hub geo-DR alias. Similar to the 'primaryConnectionString' property, but using the alias primary connection string.

For example, "[reference(....).primaryConnectionString]" works for the non-alias connection string, but "[reference(....).aliasPrimaryConnectionString]" doesn't exist.

Here's the error if you try it:

The language expression property 'aliasPrimaryConnectionString' doesn't exist, available properties are 'primaryConnectionString, secondaryConnectionString, primaryKey, secondaryKey, keyName'..

Event Hub Disaster Recovery needs manual intervention of restarting the application to start pushing messages to secondary namespace as kafka protocol prevents connections from closing after triggering failover.

It would be nice if the kafka can automatically have a new connection at the time of failover.

Currently we have a size limit on the incoming messages on an EventHub as 1MB. Some processes batch the messages before sending them, to avoid making several POST requests. Even though the batch size is small as 10, the size is unpredictable when the message is sent, as it hits the 1MB mark. Please could we remove the size limit or at least increase the limit form 1MB to 5MB/10MB?

Kafka does not have a size limit. Google has a limit of 10MB.

An Azure Event Hub emulator (preferably in a container) to do local testing

In Diagnostic setting for any resource when sending events to EventHub it allows only namespace SAS policy. It will be good if it allows EventhHub SAS Policy.

Also today the option should say “EventHub Namespace policy name” instead of “Event hub policy name” as it confuses many users.

Currently you can capture to Data Lake Gen1 or Blob but not Data Lake Gen 2. Sad times.

Please provide message and events replication for disaster recovery in the Event hub .Currently It supports metadata replication only.We have to poll stuck messages from old primary event hub.Event hub Disaster recovery involves manual steps .Please provide feature to replicated the messages also .So that this process become automatic.

Currently event capture only supports writing in avro. It would be nice to support other file formats for processing e.g JSON and parquet.

Please support ORC format for event capture. Research suggests it could be much denser for sparse data.

FYI: It is already supported by Data Factory, SQL DW/Synapse, ADLA, and Hive (via HDInsight).