Blessed templates
It is a good practice to offer some building blocks of blessed templates for the infrastructure within many organizations. A Cloud Competence Center of Excellence commonly provides these templates. By the nature of control, the focus is primarily on the Security, Architecture, and Governance part of things, compared to Engineering enablement and Operational Excellence. The organization can run the strategy “comply or explain” by providing blessed templates, this means that it is easy to comply by using the blessed templates and no questions are asked. When you do need to deviate from the blessed templates you need to explain and go through a review board. The templates are blessed in the form that the involved parties have already approved for use. Many teams will be able to run their solutions based upon these templates.
Managing changes to the blessed templates can be a bit challenging. How do we patch it efficiently on all resources during a security vulnerability? What if a product team has moved on to the next project and do not actively support the previous project? How would teams know if they need to change or re-deploy their infrastructure? The effectiveness stands or falls by the convenience for the teams to comply. How easy is it to reach teams that use your template, and how easy is it to change the template and re-deploy.
Using the modular improvements introduced in Bicep v0.4.1008 to support the Bicep registry, you can improve your support of blessed templates to your consumers and have compile-time validation to support complete CI/CD scenarios for your IAC (Infrastructure As Code). Let’s see how this impacts the ease of use and blessed templates lifecycle.
Basic Bicep deployment
To understand the value of using a Bicep registry for your templates, we first need to understand how things work without templates, as shown in the diagram below.

A relatively standard CI/CD pipeline for infrastructure written in Bicep, where you don’t use templates:
- An engineer makes a change to a Bicep file in Git
- When pushing the change, a pipeline will be automatically triggered
- The first step in the pipeline is to transpile the Bicep template into an ARM template and store that as an immutable artifact for later use
- Deploy the artifact, with environment-specific parameters, to an Azure -ResourceGroup, -Subscription, -Tenant, or -Management group
The deployment is nothing more than running an az CLI command using the “deployment group” arguments with the transpiled Bicep template file passed with the –template-file switch, as shown in the following Powershell.

Using templates
The use of templates helps you re-use definitions you already created. When deploying a web service, you always want to deploy application insights with a log-analytics workspace. Using templates this is done in multiple ways, for example, by referencing;
- a (local) folder in the same project
- a storage account in Azure
- a template-spec resource in Azure
- a module in the Bicep registry
Template-spec
Before the possibility to push Bicep templates to a Bicep registry existed, the preferred way of sharing your templates was by publishing the ARM template to a Template-Spec-Resource. Below you can see such a template-spec resource in Azure.

Notice how this can leverage RBAC (a consumer needs only read-access), versioning, and even release documentation when drilling down to a specific version. A template-spec is published using the CLI like this:

Another Bicep file can then use this Template-Spec-Resource by using “Microsoft.Resources/deployments” type with a “templateLink” property referencing the Template-Spec-Resource with a specific version as shown below:

The template link describes a full URI, including the desired resource’s name and version.
Compared to our Basic Bicep Pipeline, using the “Template Specs” helps you already to achieve a blessed template structure, having semantic versioning and release documentation together with your template definition. Template specs are already altering the CI/CD flow using the blessed template during deployment.

The maintenance, approval, and availability are taken care of by a separate team, having the ability to publish new templates and versions. The consumers only have read privileges on the template spec and can use it during the deployment of their resources.
The use of the template-spec as a URI reference, as shown in figure 3, is a bit clumsy. Due to not having IntelliSense over the Template-spec, you have to know the names of the parameters to pass them. The same applies to the output of the template spec. You also need to know the name of the parameter to retrieve it. We can improve on this using a module instead of a resource. A module can refer to a local file or a template spec. To use a template spec, use the following format:
module <symbolic-name>'ts/<alias>:<template-spec-name>:<version>' = {
When looking at the template below, you can see no direct reference to a subscription anymore:

The reference is abstracted away into a configuration file, making your definitions more readable and easier to maintain.

You place the configuration file, called called “bicepconfig.json”, in the root of your project. Here you can define the “ts” “Template Spec”. The alias in this example is ‘BlessedTemplates’. The other benefit of using the module approach is Bicep will recognize the link, start downloading the definition to your local user’s folder, and provide IntelliSense during your development. It, therefore, becomes much easier to use parameters or the outputs of the resource.
There are a few shortcomings of using “Template Specs”:
- While using template specs, you reference it by a link. This causes validation to happen at deployment time, instead of build time, which is a bit late
- The content of the template-spec is not known client-side. Looking at the ARM template, you will only see the reference to the template-spec instead of seeing the nested resources of the spec. In our example of the web services, you would only see a reference to a web service template-spec and not be aware that an application insight and a log analytics workspace would also be deployed.
Bicep registry
Using the modular improvements introduced in Bicep v0.4.1008, we can now use a Bicep module registry. This improvement enables us to publish Bicep modules to an Azure Container Registry, as shown in the following command.
az acr login --name mybicepsharedregistry.azurecr.io
bicep publish StorageAccount.bicep --target
br:mybicepsharedregistry.azurecr.io/bicep/modules/storage:0.1
First, you need to log in to the Azure Container Registry and publish the Bicep file. Next, add the configuration to the bicepconfig.json and reference the module as you did for the template spec. This time, you use the “br” keyword. This keyword helps Bicep to understand it can retrieve the modules from the bicep registry and thus enabling IntelliSense and compile-time validation.


Another significant benefit, compared to the template-spec, is that when the Bicep file is transpiled into an ARM template, you will get nested templates instead of a link to the template. A nested template explains what resources will be modified. In contrast, a template link only references a template spec that will be accessed during deployment, making it harder to understand what is happening while reviewing an artifact for deployment approval. As an example, the first screenshot below shows the use of a template spec, the second one the use of a module in the Bicep registry.


Shift left
Using either the template-spec or bicep registry will gain the Shift Left2 capability of compile-time validation instead of deploy time validation as shown below.

Because Bicep will download the ARM templates and the Bicep files to the local user’s folder, it will validate during compile-time. Compile-time validation will help you fail your pipeline in the build step before creating the immutable artifact you want to deploy to your environments.
Renovate-bot dependency automation
One of the questions we asked at the beginning of this article was how we manage changes to our blessed templates and enable our consumers to detect changes that they need to deploy quickly. We can use a dependency manager like Renovate-bot to detect new versions using semantically versioned Template-specs or Bicep registries.

Implementing the Renovate-bot will enable the following flow:
- Renovate-bot will scan the organization’s repositories for out of date dependencies
- Renovate-bot submits a Pull Request into the repositories that use the blessed templates, enabling your consumers to approve or auto-approve the Pull Request and stay secure and compliant.
- Approval will automatically trigger your CI/CD pipeline and roll out the new templates to their environment.
Configure Renovate-bot
To make use of the Renovate-bot, follow the website 3 guidelines. Renovate-bot can be integrated with industry-standard CI/CD tooling and runs on a hosted or on-premise environment. The easiest way to enable this is by installing it as a service into your GitHub account.
As a team managing the blessed templates, you want Renovate-bot to pick up on published changes. To pick up those changes you can use an example as shown below.

To always have a valid version, this script does the following:
- Get the current version from the latest Bicep registry manifest
- Increment the minor version to create a new, unused version
- Publish the Bicep with the incremented version to the registry
- Create a GitHub tag on the current SHA4 used to run the build
Now that we’ve tagged the release in GitHub, we can use the Renovate Managers5 to configure an override for bicep files. Renovate managers are like package managers. These managers know, for a specific resource (such as docker, dotnet, golang, etc.), how to determine the latest published version and compare it to the version used in the repository. Because there is no dedicated manager for Bicep, we need to configure our own using the generic regex manager.

To configure the use of the regex manager, we change the renovate.json and add a “regexManagers”:
- Configuration to match all *.bicep files. This configuration will limit this manager’s configuration to only search in the .bicep files and not any other files you have set up.
- Define your “matchStrings” regular expression. This configuration will search for the semantic versioning in the “br:****” annotation. A regular expression group ‘currentValue’ will contain the found version.
- Configure the “datasourceTemplate” so the regexmanager can compare the ‘currentValue’ to GitHub-tags. GitHub-tags is a known data source6 for renovate-bot
- Configure the GitHub repository to search for the tag in “depNameTemplate”
The result is an automatic Pull Request whenever a new blessed template publishes in a foreign repository.

Conclusion
Many organizations already have blessed templates, and their success depends on the ease of use. The blessed templates should be beneficial for the organization to make sure that all consumers work in a secure and compliant manner. Consumers should rely on the service offered by the blessed templates to update their dependencies to maintain compliance automatically. Using this provided service should not be hard. It should be a golden path to take, enabling engineering capability instead of restricting it.
Take away
- You can keep using the existing approved blessed templates with Bicep
- You can make use of IntelliSense by using bicep-configuration
- Using bicep registries enables you to have compile-time validation (which is actually transpile time)
- Use Container Registry over Template Specs to have a more explicit transpiled ARM template.
- Support all repositories with automated dependency updates on your blessed templates using Pull Request created by renovate-bot
You must be logged in to post a comment.