Spinning up 'Severless' Functions in Azure

Since I started using Azure App Services in 2016, Microsoft has seriously pushed their cloud technology, there is an absolutely massive stack to get to know. There's everything from web servers, to SQL servers, to data science toolboxes (using Jupyter Notebooks), cloud based network infrastructure, and the list goes on and on.

One of the most productive tools that we've been using over the last few years has been Azure Functions, which are Microsoft's answer to AWS serverless functions. You can write Azure functions in a number of language-frameworks, including Node.js in VS Code, but today we'll cover writing these in C# out of Visual Studio 2019.

The Basics

If you're unfamiliar with the concept of a 'severless' function, it's basically a way for you to rent a few minutes of computation time to perform a simple task. We use these to pull data from a software system, process it, and store it in our data warehouse. Conversely, we pull data out of the data warehouse with Azure Functions, and use the data to sync up two discrete software systems.

The underlying technology here is Asp.NET Core, so if you're familiar with Asp, or Azure App Services, a lot of this will be familiar to you.

Time & Atomicity

Azure functions can only run for an upper limit of 10 minutes in production, so you'll need to make sure your task can be completed in 10 minutes. Further, should the function not complete for any reason, you'll need to start over. It's important that you don't write any data to your database unless the function has run to completion. This 'all-or-nothing' approach is extremely important, because the app service may stop mid-task without warning, and the last thing you want is a half-completed update task in a production database (what a mess!).

With the Mongo libraries, for example, you'd want to use a bulk upsert wrapped in a transaction. Process the data first, then when ready, update the database. If the function app service dies in the middle of the task, nothing is written. With Transact-SQL, the approach is roughly the same, make sure your write-operation is wrapped in a transaction, and the process becomes atomic.

Getting Started

Getting going in visual studio is quite easy, you'll just want to create a new Azure Function from the project template.

The first choice to make here is how you'd like to trigger your function app. Our Azure Functions serve to automate business process and sync data from one system to another, as well as pull data for analytics. So for our purposes timer triggers will do just fine, however you can trigger these via HTTP, queue services, or a number of other approaches.

As a side note, SignalR is a framework connecting web applications to a backend without requiring the page to refresh, so it's interesting to see that we can trigger Azure Functions with it too.

 

Writing the Function App

Once you've setup the App (Visual Studio has done most of the work for us, which is always great for small teams like mine), you'll a few things already there for you.

First is the Cron timer used to trigger the app, the default one runs every 5 minutes, but you can make it run on any time cycle you can think of. The first of every month, every second Tuesday, once a day at midnight etc.  

 
You'll also notice that there's some default log information being written out to console here. This is how you'll push information out to your Azure Function console, which is very important for debugging. You can get a logstream in Azure, which is super useful if you need to see the function running in test or in prod.

Here it is running in my dev environment. Easy!
 

 

 Final Notes

Getting an Azure Function up and running takes 5 minutes. It's only a matter of just writing the code to do what you want to do. In my experience it's quite limited scope regardless of what you need it to do. To further multiply your capacity, you can use Azure's GitHub integration to setup Continuous Integration. Just point Azure at your production branch, and your Function App Service will see when you've pushed a new revision upstream.

I'll cover this in more detail next week, since CI in Azure is a bit broader in scope, and is useful beyond just Azure Functions.

 

Comments

Post a Comment

Popular posts from this blog

Creating HID Override Layouts for the Unity Input System

Image
     Unity's new(ish) Input System is a quantum leap from their old Input Manager system. You can re-bind controls, axes, and buttons from various devices with relative ease. You can broadcast events from devices and handle those events inside your unity C# scripts very easily. It is, in my estimation, a good system. It's young code, and it comes with its own set of quirks as a consequence, but the documentation is complete, if a little obfuscated by overly complex examples.  Yesterday I was able to get the Input System up and running with my HOTAS. Dapper Dino did an excellent video tutorial on YouTube. If you haven't gotten up and running with Input System yet, I strongly suggest you start here: So I was pretty pleased with myself, after porting most of the input in my project to the new system. However when I went to add yaw controls to my spaceship, there was a problem; Unity didn't detect my rudder pedals. Alright then, time to dig into the Input System document...
Read more

Building Ben Eater's 8-bit CPU Clock (part 1)

Image
Okay so, I know I said I'd cover continuous integration in Azure this week, but I've been putting off this 8-bit CPU project for too long, and today is the day. If you don't know, Ben Eater runs a fantastic YouTube channel on digital electronics. For me, spending most of my career doing high-level software engineering, this channel served as both an excellent intro into digital electronics, but also to demystify the topic entirely. Getting Started So, Ben Eater sells kits for his 8-bit CPU on his really well structured website. I'm a sucker for good documentation, and Ben Eater has aced it here. You can find his website here: https://eater.net/8bit Now, you don't need to actually buy the kit from Ben, you can see the parts list that he has kindly included for you. I'm not that ambitious, I'd rather save the time and effort and just get the kit. But if you're short on cash, you can pick up these parts on the cheap. Tools There's a number of tools you...
Read more

Managing API Keys for Serverless Functions with Azure KeyVault

Image
With my team's ever growing list of interconnected services, serverless functions, external APIs, databases, et.al., we need a way to securely manage and protect all the sensitive information contained therein. This is not an uncommon use case, and in-fact several high profile breaches over the last few years involved the storage of secure info like API keys in source control. We'll be using Azure KeyVault to manage our sensitive keys, and today we'll look at how to connect an Azure Function to an Azure Key Vault. Before we get started, this is going to be a zero code demonstration. Our development environment doesn't need access to the KeyVault for you to trigger the function, even with AuthorizationLevel.Function set. We will do the setup purely in Azure, and show you how to setup role based access. I'm going to assume you are familiar with Azure and know how to setup new services. The Microsoft documentation on this authorization scenario points at legacy access ...
Read more