Course Content

Building custom Bicep Local Deploy extensions From fundamentals to publishing your first custom-ready extension 4 Lessons

Introduction to Bicep Local Deploy extensions 20 min
Creating your first Bicep Local Deploy extension 45 min
Testing your Bicep Local Deploy extension 60 min
Documentation and publishing Bicep Local Deploy extension 35 min

Course content

This course takes you from Bicep Local Deploy fundamentals to publishing custom extensions that are production-ready. The getting started module starts with an introduction to the Bicep Local Deploy framework and its architecture. It covers how custom extensions work within Bicep's ecosystem. You'll set up your development environment and verify your installation using a quick-start project.

Next, you'll be building your first complete extension, expanding on the quick-start project. From looking at the project structure to defining type models with attributes, and lastly, implementing handlers that manage the resource lifecycle.

The advanced module lessons dive deeper into testing and documentation. You'll be using the bicep-local-docgen CLI tool to validate your documentation.

Prerequisites

Before starting this course, you should have:

C# programming: Be familiar with C# syntax, classes, and properties.
PowerShell basics: Understand PowerShell commands and scripting for running build scripts and tests.
REST APIs: Basic understanding of HTTP methods (GET, POST, PUT, DELETE) and JSON.
Command-line tools: Be comfortable running CLI utilities.

Don't worry if you're missing some recommended knowledge. The course provides explanations along the way. The required knowledge is mainly for following the code examples and completing the exercises.

What you'll learn

Understand Bicep Local Deploy architecture: Learn what custom extensions are, how they extend Bicep's capabilities, and the role of Azure.Bicep.Local.Extension framework and documentation attributes.
Build complete custom extensions: Create type models with documentation attributes, implement resource handlers, and test with Bicep files and Pester.
Generate and publish documentation: Use bicep-local-docgen CLI to generate markdown documentation and validation.

By the end of this course, you'll have a production-ready Bicep Local Deploy extension, including:

Documentation
End-to-end tests
Publishing workflow

You'll understand the complete lifecycle from initial concept to a published extension that users can consume in their Bicep templates.

Plan your migration Inventory your PowerShell DSC estate and choose a path for every artifact 7 Lessons

Introduction to migrating from PowerShell DSC 4 min
Why migrate, and when not to 8 min
Inventory what you have 12 min
Build and inventory the Bindery estate 20 min
Choose a migration path 10 min
Knowledge check 5 min
Summary 3 min

Migrate configurations and resources Walk all four migration paths against one legacy estate 10 Lessons

Introduction to migrating configurations and resources 3 min
Path A: Adapt in place 12 min
Adapt the legacy configuration 15 min
Path B: Import your compiled MOFs 12 min
Convert a MOF file 15 min
Path C: From script-based to class-based 15 min
Rewrite AppEnvironment as a class 20 min
Path D: Going native 8 min
Knowledge check 5 min
Summary 3 min

Replace the LCM Own the enforcement schedule, handle reboots, and migrate credentials 9 Lessons

Introduction to replacing the LCM 3 min
Map LCM features to Microsoft DSC 12 min
Own the enforcement schedule 10 min
Reboots and cross-node coordination 10 min
Migrate credentials 10 min
Prove equivalence end to end 20 min
Pitfalls from the field 8 min
Knowledge check 5 min
Summary and next steps 5 min

Course content

This course takes you from a working PowerShell DSC estate to a Microsoft DSC estate, without a rewrite weekend and without breaking production. The planning module starts where every real migration starts: with an honest look at what you already own. You'll learn why the trajectory points to Microsoft DSC, build an inventory that classifies every resource on a machine, and use that inventory to assign each artifact to one of four migration paths.

Next, you'll walk all four paths against the same small legacy estate. You'll adapt a Configuration script into a configuration document with the resource code untouched, rescue a compiled MOF file whose source script is long gone, rewrite a script-based resource as a class-based one that runs on PowerShell 7, and learn the signals that justify a full native rewrite.

The final module deals with the part of PowerShell DSC that isn't a file: the Local Configuration Manager. You'll map every LCM behavior your estate might depend on to its Microsoft DSC equivalent, replace scheduled enforcement with something you own, move credentials out of MOF files, and run an end-to-end equivalence check that proves all three migration routes converge on the same machine state.

Prerequisites

Before starting this course, you should have:

PowerShell DSC experience: You've written Configuration blocks, compiled MOF files, and run Start-DscConfiguration at least a few times.
Microsoft DSC fundamentals: You can read a configuration document, and you've run dsc config get, dsc config test, and dsc config set against one.
PowerShell scripting: Functions, modules, module manifests, and enough class syntax to read a [DscResource()] class.
A Windows machine: With both Windows PowerShell 5.1 and PowerShell 7+ installed, plus the dsc CLI on your path. Administrator rights are needed for the machine-level environment variable used in the exercises.

Don't worry if you're rusty on the Microsoft DSC side. Each unit explains the construct before using it, and every command in the exercises is shown in full. The PowerShell DSC knowledge is the part that's hard to substitute, because this course is about translating something you already own.

[!TIP]
If you've never applied a configuration document with Microsoft DSC, work through the getting-started chapters of The Microsoft DSC Handbook or The Fundamentals of Microsoft DSC first. This course assumes the mechanics and focuses entirely on the translation.

What you'll learn

Inventory a PowerShell DSC estate: Classify configurations, compiled MOFs, script-based resources, and class-based resources, and check what each Microsoft DSC adapter can already see.
Choose a migration path per artifact: Apply a four-path decision framework — adapt in place, import MOFs, rewrite as class-based, rewrite as native — and defend the choice with cost and payoff, not preference.
Execute each path hands-on: Translate Configuration scripts into configuration documents, convert MOF files with an import extension, and port Get/Set/Test-TargetResource trios into class-based resources.
Replace the LCM: Map enforcement modes, refresh intervals, reboot policy, partial configurations, and cross-node coordination to explicit Microsoft DSC replacements that you schedule and own.
Migrate credentials safely: Retire PSDscAllowPlainTextPassword and MOF encryption certificates in favor of secureString parameters and secret extensions.

By the end of this course, you'll have migrated a complete (if small) estate end to end, including:

A migration inventory report you can run against any machine
A configuration document that replaces a Configuration script
A class-based resource ported from a script-based one
A scheduled enforcement replacement for the LCM
An equivalence check that proves the migration didn't change behavior

You'll understand the full lifecycle from "we have hundreds of MOFs and nobody remembers who wrote them" to a reviewed, tested, version-controlled Microsoft DSC estate.

The running example

Every unit uses the same fictional estate, owned by a company called Bindery. A bindery is a workshop where loose sheets are fastened together into books, which is a fair description of what a migration does: fasten an old estate to a new engine without tearing any pages.

Bindery's operations team owns:

BinderyOps, a PowerShell module containing a script-based resource named AppEnvironment that manages a machine-level environment variable.
LegacyWebServer.ps1, a Configuration script that uses AppEnvironment alongside the built-in File resource.
output\localhost.mof, the compiled artifact that the LCM actually applies.

It's small enough to follow in a single sitting and structurally identical to the thousand-line resources in your production estate.

Getting started Learn the fundamentals and set up your development environment 4 Lessons

Introduction to Microsoft Desired State Configuration (DSC) 5 min
Installing Microsoft Desired State Configuration (DSC) 10 min
Your first resource invocation 10 min
Practicing and exploring 20 min

Course content

This course takes you from the Microsoft DSC fundamentals to practical implementation. The getting started module introduces Microsoft's Desired State Configuration (DSC) cross-platform command-line tool for declarative configuration management. You'll learn the core concepts that differentiate DSC from traditional configuration management approaches, install Microsoft DSC on your system, and get hands-on experience by invoking your first DSC resource. At the end, there's room for practical exercises and a knowledge check.

What you'll learn

Understand Microsoft's DSC architecture: Learn what Desired State Configuration is, how it's different from imperative tools, and why idempotency matters.
Install and configure Microsoft DSC: Set up Microsoft DSC on Windows, Linux, or macOS, including PATH configuration and installation.
Work with DSC resources: Discover how to explore resource schemas, understand property requirements, and invoke the core resource operations (get, set, test, and delete).
Configure your registry: Gain practical experience with the Microsoft.Windows/Registry DSC resource.

-/-

Map LCM features to Microsoft DSC

During migration planning, someone will ask about every single LCM setting. Here's the complete mapping in one place, so the answer is a row in a table rather than a research task.

LCM feature	Microsoft DSC equivalent
`ApplyOnly` mode	Run `dsc config set` once (from a pipeline, script, or by hand)
`ApplyAndMonitor` mode	Schedule `dsc config test` and alert on drift; nothing auto-corrects
`ApplyAndAutoCorrect` mode	Schedule `dsc config set`; each run converges the machine
`ConfigurationModeFrequencyMins`	The schedule you define in Task Scheduler, cron, or your orchestrator
`RefreshMode = 'Pull'` + pull server	A higher-order tool such as OpenDSC
`RebootNodeIfNeeded`	The `restartRequired()` function and restart metadata; the caller decides to reboot
`PartialConfigurations`	`Microsoft.DSC/Include` composing multiple documents
Cross-node `WaitFor*` resources	No equivalent by design; cross-node coordination belongs to your orchestration layer
`PSDscAllowPlainTextPassword`	Gone, and good riddance: `secureString` parameters and the `secret()` function
MOF encryption certificates	Not needed; documents hold no secrets when parameterized correctly
`DscLocalConfigurationManager` state	Nothing; the engine is stateless, results are returned to the caller

Read the table as three groups

The rows aren't equivalent in kind. They fall into three groups, and conflating them is what makes migration planning meetings run long.

Group 1: Direct substitutions

ApplyOnly, ApplyAndMonitor, ApplyAndAutoCorrect, and ConfigurationModeFrequencyMins all reduce to "which command, how often." There's no conceptual gap here — only a decision about where the schedule lives, which unit 3 covers.

PartialConfigurations is also a direct substitution. What the LCM assembled from multiple sources, Microsoft.DSC/Include composes explicitly in a document:

resources:
  - name: Baseline
    type: Microsoft.DSC/Include
    properties:
      configurationFile: ./baseline.dsc.config.yaml
  - name: WebRole
    type: Microsoft.DSC/Include
    properties:
      configurationFile: ./webserver-adapted.dsc.config.yaml

The difference is visibility. Partial configurations arrived from wherever the LCM was told to look, and reconstructing what a node actually ran meant querying the node. An Include is a line in a file you can read in a pull request.

Group 2: Responsibilities that moved to you

RefreshMode = 'Pull' and RebootNodeIfNeeded didn't disappear — they moved out of the engine.

Pull mode needs a higher-order tool. Today, that's OpenDSC, a community project that acts as the enforcement and orchestration layer above the engine: nodes register, receive configurations, report results.

[!IMPORTANT]
OpenDSC is community-driven open source and isn't officially supported by Microsoft. If your organization requires vendor support for the enforcement layer, decide that before decommissioning a pull server.

Reboot policy moved to the caller, which unit 4 covers in detail.

Group 3: Things that were cut, not moved

Cross-node WaitFor* resources have no equivalent, deliberately. So does LCM state.

This is the group that surprises people, because the honest answer is "that feature is gone and you need to redesign around it." Unit 4 covers what to do instead.

Three rows that change who is responsible

Three of these deserve more than a table row, because they change accountability rather than syntax:

Scheduled enforcement is now yours to own. The LCM's consistency check ran every 15 minutes whether you thought about it or not. In Microsoft DSC, if nothing invokes the engine, nothing happens. → Unit 3.

Reboots became explicit. The LCM would reboot a node mid-run when a resource requested it and RebootNodeIfNeeded was set. Microsoft DSC will never reboot anything on its own. → Unit 4.

Cross-node coordination was cut, not misplaced. WaitForAll and its siblings let one node's LCM block on another node's state. Nothing replaces this. → Unit 4.

Audit your meta-configuration

Before you decide what to replace, find out what you actually configured. On any managed node:

Get-DscLocalConfigurationManager

Record these properties for each node class in your estate:

Property	What to record	Why it matters
`ConfigurationMode`	`ApplyOnly`, `ApplyAndMonitor`, or `ApplyAndAutoCorrect`	Determines whether your replacement runs `set` or `test`
`ConfigurationModeFrequencyMins`	The interval	Your new schedule should start here, then be tuned
`RefreshMode`	`Push` or `Pull`	Pull means you need an orchestration layer before you can migrate
`RebootNodeIfNeeded`	`$true` or `$false`	`$true` means something in your estate expects automatic reboots
`PartialConfigurations`	Names and sources	Each becomes an `Include` — or reveals a source nobody knew about
`CertificateID`	A thumbprint, or empty	Non-empty means encrypted credentials are in play; see unit 5

A surprising number of estates discover at this point that half their nodes are in ApplyAndMonitor while everyone believed they were auto-correcting. Better to learn that now than during the first incident after cutover.

[!TIP]
Capture the meta-configuration for every node into a file before you change anything. It's the only record of what the old system was doing, and once the LCM is disabled, reconstructing it means reading old build scripts.
Get-DscLocalConfigurationManager |
    Select-Object ConfigurationMode, ConfigurationModeFrequencyMins,
                  RefreshMode, RebootNodeIfNeeded, CertificateID |
    Export-Csv -Path .\lcm-baseline.csv -NoTypeInformation -Append

> EOF