Context : I want to keep an eye on how many resources are currently provisioned on a specific Azure subscription and understand when we are adding or removing resources.

I recently got a LaMetric Time device which can be easily configured to poll an HTTP endpoint and it was quite a while I wanted to play a bit with Azure Durable Entities that's why I assembled this simple integration.

The architecture is pretty straightforward:

In this example I am using Azure functions to query the Azure Resource Graph to count the resources using a simple query like the following:

Resources | summarize count()

The result of the query is persisted using an Azure Durable Entity function activated via an HTTP.

The entity state looks as the following:

{
    "entityExists": true,
    "entityState": {
        "Name": null,
        "PreviousRead": 575,
        "LastRead": 571,
        "UpdatedOn": "2021-04-01T19:44:04.5683114Z"
    }
}

The output of the GET call returned to the LaMetric device is formatted this way:

{
    "frames": [
        {
            "text": "-4", // increment
            "icon": 124 // up or down icon
        },
        {
            "text": "571", // last read
            "icon": 37287 // azure icon
        }
    ]
}

Here the final result.

Arcus.Observability is used to simplify the integration with Application Insights.

• Entities stateful addressable singletons.

• Entities can be called either from durable clients (IDurableEntityClient) or from durable orchestrations (IDurableOrchestrationClient).

• Entities guarantee that requests are processed in series.

• Entities are triggered via 'control' queues (default 4) prefixed with the {task-hub-name}.

  • Only a single Azure Function instance can read from a single queue at a certain time. This is guarantee by the blob leases approach of the durable framework.
  • from a scale prospective, 4 queues means maximum 4 azure function instances. In my case 1 function instance interacted with two queues.

  

  • the Functions scale controller add/remove instances according to the queue latency for peeking messages.

  

• Data persistence for Durable Functions.

• Performance and scale of durable functions.

The Entity instances are persisted in the 'Instances' table with the key @{entityname}@{entityid}

History table contains the events occurred for every instance. There is one table per "hubname".