Slightly OT: Proposal to provide Virtual Resource Groups on Azure

[SetTrend]

For being able to test ARM templates, Bicep etc, it would be very feasible if Azure provided Resource Groups that don't serve real resources, but only mock resources. Then, ARM template deployment could be tested without any Azure computing resources being drained or charges/fees to be payed.

See here for the full idea on Azure Feedback.

[brwilkinson]

That's an interesting idea. Also it would be a good goal to be able to at least plan do this, at some point in the future.

One of the challenges, which I think is exposed with the whatif command, is that providers and api's were not originally written with this in mind.

https://docs.microsoft.com/en-us/azure/azure-resource-manager/bicep/deploy-what-if?tabs=azure-powershell%2CCLI

If you are familiar with the whatif results, you would know that there is "noise". Where api definitions are not always accurate.

Also another challenge/consideration is the stacking of resources, which are the layering of dependencies. Especially this is a challenge, since many Resource Providers are developed in somewhat isolation.

For the time being, have you been able to acquire a developer subscription? Some have $x per month to test and deploy resources. I think between $50 to $150. One way to get one is via Microsoft Certified Trainer MCT... However there are other methods.

I hope we get more discussion and feedback on this topic.

[SetTrend]

Thanks, @brwilkinson!

Yes, I believe the issue with the whatIf option is that it only spans a single Bicep operation. And it's not very visual.

I received a comment at StackOverflow, suggesting to split my endeavour to deploy a resource group along with its resources (a common scenario) into two distinct deployment operations. (A suggestion that I don't favor at all. I expect my endeavour to be atomic, to be run from a single Bicep deployment operation.)

FWIW, I running an Visual Studio Enterprise subscription for decades.

[SetTrend]

I understand it's expensive to implement and that it hasn't been considered by Microsoft when originally launching Azure.

Yet, with the suggested approach, people could learn Azure in their individual ways, may it be via portal, ARM templates or Bicep – without having to worry about the financial implications.

In the future, there may perhaps be an option to convert the proposed mock resource groups into "real" resource groups having actual computing power enabled (e.g. by simply renaming a resource group from "__RG" to "RG" — see original suggestion above).

This would be ideal for complex deployments, where administrators could test their deployment scripts first, verify the implications online, and then alter their scripts for creating real resource groups. That'd be great for CI/CD scripts, creating several stages for the same project.

[brwilkinson]

One other aspect in regards to learning that I think is a home run... Is MSLearn and the Sandboxes there.

As you do the Modules, you spin up a REAL Azure sandbox resource group (at no cost) you get access to the Portal or Cloud shell, for the exercises, then it expires at the end of the labs.

Such a great experience and, this bypasses the cost issue and also the requirements for having any Credit card, in a way that is ideal.

[SetTrend]

Sounds great. The only flaw I assume is that – as far as I can see – with MS Learn you're not free to deploy complex, enterprise scenarios.

You know, my current customer is government IT, with hundreds of projects to deploy. They may be tempted to use ARM template specs or Bicep modules from their registry etc. Plus, we're talking about huge amount of resources here. Despite the fact that there's a WhatIf option (which would probably only represent a fraction of the deployment) it's very hard to find mistakes in a textual output. I suppose it's easier to have a good look monitoring the result on the Azure portal.

Loops, for instance. In Germany, the decimal point is the comma. Given that we introduce a number parameter that's fed with a German 15,000 value (which is 15), I suppose it's hard to find the mistake in a large WhatIf output. However, I assume it's easy to grasp from the graphical output of the Azure portal output.

You know, in Germany we have a common proverb, coined by a famous soccer coach: "The truth is in the game." ("Die Wahrheit liegt auf dem Platz.") That's why I believe the Azure portal is the right place the check the outcome of a Bicep/ARM deployment.

Yet, your suggestion enlightened me to update my suggestion: Instead of using a fix resource group prefix (although I still prefer that approach as a resource group's dummy character becomes obvious right from the name), there could be a mock location. If the user selected mock as the resource location, that would lead to non-functional mock resources.