Deploying Python Function Apps through Bicep

[Zerthick]

Hello!

I am attempting to replicate the deployment process described here https://docs.microsoft.com/en-us/azure/azure-functions/functions-reference-python?tabs=azurecli-linux%2Capplication-level using bicep but have hit a few roadblocks.

• How do I define the equivalent hosts.json settings within the Microsoft.Web/sites resource? Specifically I need to specify the extensionBundle version
• How do I ingest the python package dependencies defined in requirements.txt as part of the site deployment process?

Thank you!

[brwilkinson]

Bicep is great for creating Azure Resource Types, which are known as Resource Providers. E,g, such as a ServerFarm (App Service Plan) or a Website (App Service).

resource FARM 'Microsoft.Web/serverfarms@2021-02-01' existing = {
...
}

resource WS 'Microsoft.Web/sites@2021-01-01' = {
...
}

These are pre-requisite to deploying Applications or in this case WebSite Content AKA your Function app, which includes those files that you referenced.

I would recommend:

• To create a function instead of Bicep.exe try out Func.exe
  https://docs.microsoft.com/en-us/azure/azure-functions/functions-run-local?tabs=v4%2Cwindows%2Ccsharp%2Cportal%2Cbash

• Or Try the VSCode extension for managing functions.
  https://marketplace.visualstudio.com/items?itemName=ms-azuretools.vscode-azurefunctions

They will allow you to create and deploy function apps from command line and templates.

You will be creating some files on your local filesystem, then you have the ability to publish them, via a zipdeploy to the app service. This should get you started and also give you the ability to write your custom code within the function.

By the way, you can also create the Function app itself with these above tools, which will setup a variety of App Config settings for the language of your choice depending on the template that you selected.

Once you have a working solution, it's easier to reverse engineer some of the bits into a Bicep Template. One method that I like is to get the ResourceID of the function, then in VSCode, you can select the command palatte, then "Bicep: Insert Resource", then you paste in the resourceID and it will paste in the Bicep code into your bicep file in vscode.

• The setup experience in the portal is also a good way create your first function, if you are not interested in using func or vscode function extension.

[Zerthick]

Hello @brwilkinson, thanks for the response! I was able to get my function app deployed using func as you describe, what I'm trying to figure out is if the use of func is necessary to package and deploy my python code to the function or if the entire deployment lifecycle can be managed fully in bicep. I was able to get close with something like the following:

resource FARM 'Microsoft.Web/serverfarms@2021-02-01' existing = {
...
}

resource functionApp 'Microsoft.Web/sites@2021-01-01' = {
...
}

resource function 'Microsoft.Web/sites/functions@2021-02-01' = {
  name: 'myFunction'
  parent: functionApp
  properties: {
    config: {
      scriptFile: 'main.py'
      entryPoint: 'main'
      bindings: [
      ...
      ]
    }
    files: [
      {
      'main.py': 'path/to/main.py'
      }
    ]
  }
}

This correctly deployed the function and set up it's bindings, and copied the main.py script to it. However it does not install the function dependencies or copy the host.json file to the site resource. My question is if I need to rely on func to do that, or is there a way to have that functionality within the bicep script itself?

[kevball2]

The Bicep templates can be used to deploy the required Azure infrastructure resources for your function while func.exe, Azure cli, or Visual Studio Code extensions could be used to deploy your code in a separate step.

Azure Pipelines or GitHub workflows would be the ideal place to bring these steps together.

See the sample deployment below.

https://docs.microsoft.com/en-us/azure/azure-functions/functions-how-to-github-actions?tabs=python

you could extend this pipeline to include the bicep files that you have created to deploy the function infrastructure. This would be a move advanced setup. If you are looking to get started with functions the tooling in Visual Studio Code is excellent for updating your functions in Azure.

[Zerthick]

@kevball2 Yes, that was what I was trying to determine, if I needed another step in my Azure Pipeline using something like func to deploy the code or if it could all be done in bicep.

Thanks!

[brwilkinson]

code release is generally a separate process. be aware that often you may need to set app config settings with your code release. It is possible to do this via ADO task or Github actions tasks, however then be aware that redeploying the Bicep may wipe out those app config settings.

There is a workaround for that here, allowing you to read and union current app settings when you deploy.
https://github.com/brwilkinson/AzureDeploymentFramework/blob/main/ADF/bicep/AppServiceFunction.bicep#L83

Also now that you have used func to deploy, you might want to export your Bicep template via "Bicep: Insert Resource" and review the settings compared to your original Bicep template and merge back in any extra settings that func put there for you into your Bicep. Different types of apps/functions have different settings between: Windows/Linux, then Node/PowerShell/dotnet/python etc.