Build image with containerised self-hosted Azure DevOps agent and private Azure Container Registry
Introduction
Containerization has revolutionized the way we deploy and manage software. By packaging applications and their dependencies into a container, we can ensure that our software will run consistently across different environments. This is particularly useful when it comes to setting up self-hosted agents for Azure Pipelines.
A self-hosted agent is an agent that you set up and manage on your own to run jobs. You can use self-hosted agents in Azure Pipelines or Git Hub Actions. Self-hosted agents give you more control to install dependent software needed for your builds and deployments.
Why use containerised self-hosted agent
As we have seen the benefit of running the workload in container including agility, portability, high density and more. By running your self-hosted agent in a container, you can take advantage of the benefits of containerization.
- Agility: you can easily spin up new instances of your agent as needed, without having to worry about manually installing dependencies or configuring the environment. This can save you time and effort.
- Portable: Each engineering team can have autonomy and their own build agent in manage their pipeline. This help ensure that team’s builds and deployments are consistent.
- High density: Running your self-hosted agent in a container can make it easier to manage your infrastructure. Since containers are lightweight and portable, you can easily move your agents between different hosts as needed. This can help you optimize resource usage and reduce costs.
What is the challenge of using containerised self-hosted agent?
As part of CI/CD building workload container images is a key capability. It is possible to run Docker tasks on containerised selfhost agent however it does require access to the docker daemon at host level. This has critical security concerns.
A solution is to use Azure Container Registry (ACR) Task. With ACR Tasks, you can use Azure Container Registry to build container images for various platforms in the cloud. ACR Tasks also lets you automate the process of updating your Docker containers with OS and framework patches.
By default, Azure container registry accepts connections over the internet from hosts on any network. This opens an attack surface, and it is not normally in acceptable in enterprise environment. ACR Task use an agent running on Azure and required appropriated permission to able communicate with the Azure Container registry.
Proposed solution
There are a wide range of options to run containerized self-hosted agents on Azure from Azure Kubernetes Service to Azure Container App. This proof of concept Demonstrates the building of container images using self-hosted agent using Azure Container Instances with Azure Container Registry with private access only.
Architecture:
Architecture Diagram
Proof of concept:
The proof of concept is all written in Bicep and can be found at GitHub repo (https://github.com/duongthaiha/containerised-selfhost-agent,). It provision the following:
- Provision a virtual network two subnets, one for Azure Container Instance and the other for Azure Container Registry
- Private DNS zone and entries
- Provision Azure Container Registry with private endpoint
- Disable public access to Azure Container Registry
- Provision Azure DevOps self-hosted agent in Azure Container Instances
- Provision an Azure Container Registry Task to build a container image.
- Custom role definition and role assignment following minimum privilege principles.
Prerequisite:
ADO Organization and a PAT token for more details permission required please refer to Run a self-hosted agent in Docker - Azure Pipelines | Microsoft Learn
GitHub Repo and Token for more detail permission required please refer to Managing your personal access tokens - GitHub Docs
Deployment Steps:
1) Update the parameter file to include the correct parameters:
2) Run the command to deploy the resource:
3) To start the image build process, we can connect to self-host agent using Connect capability in Azure Container Registry
4) Run command to trigger build task.
5 ) From figure above you can see that self-host agent has successfully build the image and uploaded the image into the Azure Container Registry
Summary:
If you are considering setting up a self-hosted agent for Azure Pipelines, it’s definitely worth considering running it in a container and using an Azure Container Registry with public access disabled for building images with ACR Tasks.
Related Content:
Run a self-hosted agent in Docker - Azure Pipelines | Microsoft Learn
Managing your personal access tokens - GitHub Docs
Run a self-hosted agent in Docker - Azure Pipelines | Microsoft Learn
ACR Tasks overview - Azure Container Registry | Microsoft Learn
Published on:
Learn moreRelated posts
Announcing: Dynamic Data Masking for Azure Cosmos DB (Preview)
Today marks a big step forward with the public preview of Dynamic Data Masking (DDM) for Azure Cosmos DB. This feature helps organizations pro...
Use Azure SRE Agent with Azure Cosmos DB: Smarter Diagnostics for Your Applications
We’re excited to announce the Azure Cosmos DB SRE Agent built on Azure SRE Agent; a new capability designed to simplify troubleshooting and im...
General Availability: Priority-Based Execution in Azure Cosmos DB
Have you ever faced a situation where two different workloads share the same container, and one ends up slowing down the other? This is a comm...
Announcing Preview of Online Copy Jobs in Azure Cosmos DB: Migrate Data with Minimal Downtime!
We are excited to announce the preview of Online Copy Jobs, a powerful new feature designed to make data migration between containers seamless...
Azure Developer CLI (azd) Nov 2025 – Container Apps (GA), Layered Provisioning (Beta), Extension Framework, and Aspire 13
This post announces the November release of the Azure Developer CLI (`azd`). The post Azure Developer CLI (azd) Nov 2025 – Container App...
Announced at Ignite 2025: Azure DocumentDB, MCP Toolkit, Fleet Analytics, and more!
Microsoft Ignite 2025 kicked off with a wave of announcements for Azure Cosmos DB and Azure DocumentDB, setting the tone for a week of innovat...
Automating Microsoft Fabric Workspace Creation with Azure DevOps Pipelines
In today’s fast-paced analytics landscape, Microsoft Fabric has become the leader of enterprise BI implementations, one of the fundamental con...
New T-SQL AI Features are now in Public Preview for Azure SQL and SQL database in Microsoft Fabric
At the start of this year, we released a new set of T-SQL AI features for embedding your relational data for AI applications. Today, we have b...