Azure ARM模板中concat与resourceId函数的区别及dependsOn场景选型

Let's break this down clearly—first the core differences between the two functions, then which one you should use for dependsOn.

Key Differences Between concat() and resourceId() #

• concat(): This is a generic string-building tool. It takes whatever strings (or values that can be converted to strings) you pass and glues them together. For example, concat('Microsoft.Compute/availabilitySets/', variables('availabilitySetName')) will spit out a string like Microsoft.Compute/availabilitySets/myAvailSet—but it doesn't care if that resource actually exists in your template or deployment scope, and it ignores context like subscription ID or resource group name. It's just string concatenation, no more, no less.
• resourceId(): This is a purpose-built template function made to generate valid, fully qualified resource IDs. It automatically injects all the necessary scope details (subscription ID, resource group name) based on your deployment context, and it validates that the resource type and name match a resource in your template or target scope. For example, it'll return something like /subscriptions/<your-sub-id>/resourceGroups/<your-rg>/providers/Microsoft.Compute/availabilitySets/myAvailSet. It's aware of the deployment's structure, making it far more reliable.

Which dependsOn Syntax Should You Pick? #

Always use resourceId() for the dependsOn property. Here's why:

1. Scope handling: If you're working with nested templates, cross-resource-group deployments, or any scenario beyond a single simple resource group, concat() will generate an incomplete, unqualified string that ARM can't properly resolve. resourceId() automatically adapts to these complex scopes.
2. Built-in validation: ARM will check that the ID from resourceId() points to a valid resource in your deployment. If you mistype the resource name or type, you'll get an immediate deployment error—way better than a silent failure where your dependency isn't actually enforced.
3. Future-proofing: Microsoft occasionally updates resource ID formats or adds new scoping features. Using resourceId() ensures your template stays compatible without manual tweaks, whereas concat() would require you to rewrite the string structure every time something changes.

To be clear: While concat() might work in super simple single-resource-group deployments, it's fragile. resourceId() is the intended, reliable way to reference resources for dependencies in ARM templates.

内容的提问来源于stack exchange，提问作者Pallab