Autoscaling an Azure Function in a VNet

Question

I have an azure function in an S1 app service plan.

The function and its storage account are in a vnet, meaning the table storage is not publicly accessible.

I can manually scale the function via the app service plan scale out section and the function instances increase as expected.

If i try to use the autoscale option in the same section, the scaling saves, but the scaling that is defined never occurs and there is no logging to suggest the scaling evaluation is taking place.

Should autoscaling work in this vnet scenario?

Yes Auto-scaling should work in this scenario . I have some follow up questions on this. On Which particular metrics you have enabled the auto-scale option. Also just checking if you have anything under Run History ? — DixitArora-MSFT
Hi, I was running it on a schedule as I wanted to ensure the issue wasnt around access to resource metrics. There were no entries in the Run History and that screen didnt show the current number of running instances either. — matt

matt matt · Accepted Answer · 2020-02-03T12:00:44

I ended up using a premium plan to get the behaviour I wanted (scaling out as a queue grows in size).

As well as upgrading to a premium plan I had to:

Upgrade my ARM template for "Microsoft.Web/sites/config" to use apiVersion "2019-08-1"
In this template set functionsRuntimeScaleMonitoringEnabled to true (as outlined in https://docs.microsoft.com/en-us/azure/azure-functions/functions-networking-options)
Upgrade my storage extension package ("Microsoft.Azure.WebJobs.Extensions.Storage") to version "3.0.10"

Now the app will scale as the queues that the app is monitoring grow.

Autoscaling an Azure Function in a VNet

1 Answers