Scaling # of parallel instances for a queue-triggered azure function

Question

I’m trying to better understand how azure function scaling works under a consumption plan.

I have been testing an app that inserts 1000 messages in a storage queue, which triggers an azure function written in C#.

The function downloads a file and performs some processing on it. It takes around 20 seconds to complete each request.

Running the function on a consumption plan, I am seeing it take upwards of 10 minutes to fully drain the queue. Configuring each function instance to process only one queue message at a time, I would have expected 1000 instances to be launched for 1000 messages, but that does not appear to be the case.

Using the Live Metrics Stream on App Insights, I never saw the # of running instances go over 30.

Is this sort of throughput expected? For reference, here is my queue configuration in the function’s host.json:

 "queues": {
    "maxPollingInterval": 2000,
    "visibilityTimeout": "00:00:30",
    "batchSize": 1,
    "maxDequeueCount": 5,
    "newBatchThreshold": 1
  }

Any information that could be provided on the scaling capabilities, or suggestions to achieve better throughput here would be appreciated.

Answer 1

Firstly, we can specify the number of queue messages to retrieve and process in parallel (per job function) by setting batchSize in host.json. If you want only single queue message to be retrieved and processed, as you did, you can set batchSize to 1.

Besides, when we're using a Consumption plan, instances of the Azure Functions host are dynamically added and removed based on the number of incoming events. During it scales, some queue messages have be processed and removed from queue, so it does not scale to 1000 instances. Please refer to Azure Functions Consumption and how it works for detailed information.