Complete example to deploy SelfHosted AzureDevops Agents on Container Apps or Kubernetes

This example deploys a full environment to support azure devops agents. For each new pipeline, one agent container will be instanciated, and then removed once the pipeline ends.

Note: If there is no agent in the pool, the pipeline will fail. To workaround this limitation, a "placeholder" agent must be configured in order to always have 1 agent in the pool. This agent will be a "fake" agent as it will always be marked as offline.

Pre-requisites

1. Create an agent pool and save its name and id.
   Note: the pool ID corresponds to the queueId from the url when a pool is selected in the browser.
   Example: https://dev.azure.com/myOrg/myProject/_settings/agentqueues?queueId=11&view=jobs, the pool ID is 11
2. Create a PAT (Personal Access Token) with the following permissions:
   • Agent Pools: Read & manage to be able to register / unregister agents
   • Build: Read for Keda to be able to get the queue of pending builds
3. Configure the max number of parallel jobs in Azure Devops so that your ACA does scale accordingly

Deploy the solution

This repositories has 3 implementations to create self-hosted runners:

• based on Azure Container Apps:
  • arm
  • terraform
• based on Azure Kubernetes Services with Keda

Create a "placeholder" agent

There are 2 possibilities to deploy a placeholder agent:

• using the provided terraform code, the following commands will, create a VM, register it as a devops agent and finally remove it (leaving the agent in the azure devops pool):

  cd placeholder-agent
  az login
  terraform init
  terraform apply -auto-approve
  # Wait for the agent to appear in the pool before running destroy
  terraform destroy -auto-approve
  cd -

• From an azure devops agent pool, click on "New Agent" and follow the instructions. You will need a temporary VM to install the agent. The VM might be deleted once the agent has been registered in the pool.

References

• Container Apps ARM Template
• Docker Agents for Azure Devops
• Keda for Azure Pipelines
• Keda
• Terraform
• Container Apps