Very high number of Throttled Requests in Event Hub - how can we debug?

Hi @cdom22 ,
A high number of Throttled Requests in Azure Event Hubs typically indicates that the workload is exceeding the configured capacity limits. Even if the observed message rate appears low, throttling can still occur due to factors such as message size, shared namespace usage, uneven partition distribution, or burst traffic patterns.

In Azure Event Hubs, throughput is governed by Throughput Units (TUs) in Standard tier.

• Each TU provides:
• Ingress: Up to 1 MB/sec or 1,000 events/sec
  • Egress: Up to 2 MB/sec or 4,096 events/sec

Throttling occurs when these limits are exceeded. However, this may happen even when message count appears low due to the following reasons:

1. Throughput is based on data size, not just message count

Even at ~100 messages/sec, if message payload size is large, the MB/sec limit may be exceeded, leading to throttling.
2. Throughput Units are shared at namespace level

All Event Hubs within a namespace share the same TU capacity.
Other producers/consumers or hubs in the same namespace may consume capacity, contributing to throttled requests.
3. Uneven partition distribution (hot partitions)

If events are not evenly distributed across partitions, specific partitions may become overloaded, causing throttling even when overall traffic is low.
4. Burst traffic patterns

Average throughput may appear low, but short spikes in traffic can temporarily exceed the allocated capacity and trigger throttling.
5. Excessive operations and concurrent requests

High frequency of operations such as metadata calls or a large number of concurrent receive/send requests can also result in throttling behavior.

To mitigate and prevent throttling, consider the following steps:

1. Scale Throughput Units

Increase TU capacity or enable Auto-Inflate to automatically scale based on workload.
This is the primary and most effective mitigation approach.
2. Optimize message size and batching

• Reduce payload size where possible
• Batch smaller messages efficiently

3. Ensure even partition distribution

• Use appropriate partition keys
• Avoid concentrating traffic on specific partitions

4. Monitor the correct metrics

Focus on:

• Incoming/Outgoing Bytes
• Throttled Requests
• Incoming Messages vs Data volume

This helps identify whether throttling is due to size or rate.
5. Review namespace-level usage

Check if other workloads sharing the same namespace are consuming throughput capacity.

Handle throttling gracefully by implementing retry with exponential backoff to accommodate temporary service limits.
Microsoft References Links:

• Azure Event Hubs Scalability and Throughput Units [learn.microsoft.com]
• Event Hubs Quotas and Limits [docs.azure.cn]
• Monitor Azure Event Hubs
• Azure Event Hubs messaging exceptions (ServerBusy / throttling)
• Automatically scale throughput units (Auto-Inflate)

Hope this helps. If you have any follow-up questions, please let me know. I would be happy to help.

Hi @cdom22 ,
I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.

Hi,

The answer didn't really address the original question. We solved the problem by shutting down clients, but it's kinda bad we could not pinpoint the exact process that was responsible for the request spike.