Introduction to integrating Azure Functions & Flow with Dynamics 365

I haven’t paid much attention to what is happening in the Azure space (Functions, Flow, Logic Apps etc.), because I was under the impression that it is a daunting task to setup the integration i.e. Azure AD registration, getting tokens, Auth header and the whole shebang.

As a beginner trying to understand the stack and how to integrate the various applications, I have been postponing exploring this due to the boiler-plate involved setting this up. But then, I read this post from Yaniv Arditi: Execute a Recurring Job in Microsoft Dynamics 365 with Azure Scheduler. Things started clicking, and I decided to spend some days exploring the Functions & Flow.

I started with a simple use case: As a Dynamics 365 Customer Engagement administrator, I need the ability to do some simple tasks from my mobile during commute. Flow button fits this requirement perfectly. The scenario I looked into solving is, how to manage the Dynamics 365 Customer Engagement trace log settings from Flow app on my mobile, in case I get a call about a plugin error on my way to work, and need the logs waiting for me, when I get to work.

As I wanted to get a working application as fast as possible, I did not start writing the Functions code from Visual Studio. Instead, I tested my code from LINQPad as it is easier to import Nuget packages and also get Intellisense (Premium version). If you want to do execute Azure Functions locally, read Azure Functions Tools for Visual Studio on docs site. I did install and play with it, once I got completed the Flow+Function integration. Also, when you install the Azure Functions Tools for Visual Studio you get the capability to run and debug the functions locally. How awesome is that !

There are two minor annoyances that I encountered with Visual Studio development locally:

1. There is no Intellisense for csx files. Hopefully this will be fixed soon. The suggested approach in the mean time appears to be “Pre-compiled Azure Functions“. But, I did not try in this exploration phase. It also improves the Function execution time from cold start.
2. I had to install the Nuget packages locally using Install-Package even though these were specified on project.json. I could not debug the Azure Functions locally without this, as the Nuget restore did not seem to happen automatically on build.

Now, I will now specify the steps involved in creating the Azure Flow button to update the Trace Log setting in Dynamics 365 Customer Engagement.

Step 1: Head to the Azure Portal (https://portal.azure.com/).

Step 2: Search for Functions App, select the row that says “Function App” and click on create from the right most pane.

Step 3: Specify the Functions App name and click “Create”.

Step 4: Navigate to the newly created Function App from the notification area. It is also good to “Pin to dashboard” for easier access next time you login to the portal.

Step 5: Click on the “Application Settings” link from the initial Functions App screen.

Step 6: Choose the Platform as 64-bit. I got compilation errors with CrmSdk nuget packages when this was set to 32-bit. You will also have to add the connection string to your CRM instance. The connection string name that I have specified is “CRM”. You may want to make this bit more descriptive.

Step 7: Now is the exciting part. Click on the “+” button and then click on the “Custom Function” link.

Step 8: This new function will execute on a HTTP trigger and coded using C#.

Step 9: After this, I sporadically experienced a blank right hand pane with nothing in it. If this happens, simply do a page refresh and repeat steps 6-8. If everything goes well, you should see this screen. I left the Authorize level as “Function” which means that the Auth key needs to be in the URL for invocation.

Step 10: You are now presented with some quick start code. Click on the “View Files” pane, which is collapsed on the right hand side.

Step 11: Click on “Add” and enter the file name as “project.json”

Step 12: Paste the following JSON into the “project.json” file and press “Enter”. Now paste the below JSON for retrieving the CRM SDK assemblies from Nuget and press “Save”. The Nuget packages should begin to download.

{
  "frameworks": {
    "net46":{
      "dependencies": {
        "Microsoft.CrmSdk.CoreAssemblies": "9.0.0.7",
        "Microsoft.CrmSdk.XrmTooling.CoreAssembly": "9.0.0.7"
      }
    }
   }
}

Step 13: Now open the “run.csx” file, paste in the follow code and save.

using System.Net;
using System.Configuration;
using Microsoft.Xrm.Sdk;
using Microsoft.Xrm.Sdk.Query;
using Microsoft.Crm.Sdk.Messages;
using Microsoft.Xrm.Sdk.Client;
using Microsoft.Xrm.Tooling.Connector;

public static async Task<HttpResponseMessage> Run(HttpRequestMessage req, TraceWriter log)
{
    log.Info("C# HTTP trigger function processed a request.");
    var queryParameters = req.GetQueryNameValuePairs();

    // Get request body
    dynamic data = await req.Content.ReadAsAsync<object>();
    string traceLogLevel = data?.traceloglevel;
    
    var client = new CrmServiceClient(ConfigurationManager.ConnectionStrings["CRM"].ConnectionString);
	var organizationSetting = client.RetrieveMultiple(new FetchExpression("<fetch><entity name='organization'><attribute name='plugintracelogsetting' /></entity></fetch>")).Entities.First();
	var oldTraceLogValue = (TraceLog)organizationSetting.GetAttributeValue<OptionSetValue>("plugintracelogsetting").Value;
    var newTraceLogValue = (int)Enum.Parse(typeof(TraceLog), traceLogLevel);
	organizationSetting["plugintracelogsetting"] = new OptionSetValue(newTraceLogValue);
	client.Update(organizationSetting);
    if(oldTraceLogValue.ToString() == traceLogLevel)
    {
        return req.CreateResponse(HttpStatusCode.OK, $"TraceLog Level has not changed from {traceLogLevel}. No update.");
    }
    return traceLogLevel == null
        ? req.CreateResponse(HttpStatusCode.BadRequest, "TraceLog Level not found on the query string or in the request body")
        : req.CreateResponse(HttpStatusCode.OK, $"Trace Log updated from {oldTraceLogValue} to {traceLogLevel}");
}

enum TraceLog{
	Off,
	Exception,
	All
}

Step 14: You can now execute this function by click “Run” and using the JSON in the screenshot for the POST body. The “traceloglevel” can be one of three values: Off, Exception and All.

As you can see the function:

1. Connected to the organization specified in the Application Settings using the connection string
2. Retrieved the current trace setting and updated it, if there is a change, using the SDK
3. Returned the response as text/plain.

If you want to execute the same using Postman or Fiddler, you can grab the Function URL as well. Note that the AuthToken is in the URL.

Step 15: Since I am going to do an update, I don’t want the function call to trigger the change. So, just turn off “GET” and save. This means that the “traceloglevel” will only be updated on a “POST”, and not on a “GET” with query string.

Step 16: Now it is time to export the API definition JSON for consumption by Flow.

Step 17: Choose “Function Preview” as the API Definition Key and then click “Generate API Definition Template” button to generate the Swagger JSON.

Step 18: Now click on the “Authenticate” button and enter the function auth key (see Step 14) in the API Key textbox and click on “Authenticate” button in the dialog box.

You should see a green tick next to the apiKeyQuery. This means that the key has been accepted.

Step 19: Now it is time to add the post body structure to the Swagger JSON. I used the Swagger editor to play around with the schema and understand how this works. Thank you Nishant Rana for this tip.

You should now able able to POST to this function easily and inspect the responses.

Step 20: Now click on the “Export to PowerApps+Flow” button and then on the “Download” button. You should now be prompted to save ApiDef.json into your file system.

Step 21: Now it is time to navigate to Flow

Step 22: You can now create a custom connector to hookup Function and Flow.

Step 23: It is now time to import the Swagger JSON file from Step 20. Choose “Create custom connector” and then “Import an OpenAPI file”. In this dialog box, choose the Swagger JSON file from Step 20.

Step 24: Specify the details about the custom connector. This will be used to later search the connector when you build the Flow.

Step 25: Just click next, as the API key will be specified on the connection, not on the connector. The URL query string parameter is “code”.

Step 26: Since I just have only “ModifyTraceLogSetting” action, this is the only one that shows up. If you have multiple functions on the Functions app, multiple operations should be displayed on this screen.

Step 27: If you navigate down, you can see that the connector has picked up the message body that is to be sent with the POST.

Step 28: If you click on the “traceloglevel” parameter, you see see details about the POST body.

Step 29: This is the time to create the connection that will be used by the connector.

Step 30: Enter the Function API key that you got from Step 14. This will be used to invoke the Function.

Step 31: The connection does not show up straight away. You will have to click on the little refresh icon that is to the right of the Connections section. You can now test the connection by clicking the “Test Operation” button, and choosing the parameter value for “traceloglevel” that will be sent with the POST body. You can also see the live response from the Function on this screen.

Step 32: Once you have saved your connector, you will see something like this below, on the list of custom connectors.

Step 33: Now is the time to create the Flow. Choose My Flows -> Create from blank -> Search hundreds of connectors and triggers

Step 34: Enter the Flow name and since this will be invoked from the Flow app on mobile, choose “Flow button for mobile” as the connector.

Step 35: The Flow button will be obviously triggered manually.

Step 36: When the user clicks the Flow button, it is time to grab the input, which in the case will the Trace Log Level setting. Choose “Add a list of options” and also a name for the input.

Step 37: You don’t want the user to enter free-text or numbers, hence you present a list of options from which the user will choose one.

Step 38: After clicking “Add an action”, you can now choose the custom connector that you created. Search and locate your custom connector.

Step 39: Flow has magically populated the actions that are exposed by this connector. In this case there is only one action to modify the Trace Log setting.

Step 40: In this step you don’t want to choose a value at design time, rather map the user entered value to the custom connector. So, choose “Enter custom value”.

Step 41: The name of the input in Step 37 is “Trace Level”, so choose this value as the binding value that will be used in the custom connector.

Step 42: In this case, I have a simple action. I just want to receive mobile notification.

Step 43: I just want to receive a notification on my mobile, since I have Flow app installed. When my custom connector calls the function that updates the trace log level, the response text that is returned by the function comes through on the body on the Flow app.

This text is displayed as a notification. If you have a JSON returned by the Function and Flow app, you have to use the parseJSON manipulation to grab the right property. In this case, it is not required as the response is plaintext.

Step 44: When the Flow design is complete it should look like this.

Step 45: You can run the Flow from either the Flow app on mobile or right from here. I click “Run Now” to check if everything is OK. You can also specify the “Trace Level” here that will be passed to the Function.

Step 46: I can check the status of the Flow easily. The cool thing about this screen it that it logs so much information that is useful while troubleshooting what went wrong.

I can also invoke this Flow on my mobile, using the Flow App. I get a native notification when the Flow completes.

What’s next

While I was experimenting with Flow and Function, I wanted to test integration between Slack and Dynamics 365. For proof of concept, I am running a custom command (“/recordcount”) on Slack channel to retrieve records from Dynamics 365.

I will blog about this next.

Conclusion: I am really excited about the future of Flow & Functions and what this brings to the table for both developers, who want to get their hands dirty and power-users, who want something that they can hook up easily without writing any code.

If you have any feedback, suggestions or errors in this post, please comment below, so that I can learn and improve.