Azure DevOps Agents on AKS with KEDA Autoscaling

This repository provides a Helm chart for deploying self-hosted Azure DevOps agents on Azure Kubernetes Service (AKS) with KEDA-based autoscaling. The solution ensures efficient resource utilization by scaling the agents dynamically based on pending jobs in the Azure DevOps pipeline.

Features

• Docker-based agent deployment — Ubuntu 24.04 LTS base image with Azure Pipelines agent
• Helm chart for easy deployment — deploy with configurable parameters via values.yaml
• KEDA Autoscaling — dynamically scales agents based on Azure DevOps pipeline queue length
• Secure authentication — uses Kubernetes Secrets for storing Azure DevOps PAT
• Customizable configurations — agent pool, image version, and scaling limits are all configurable

Repository Structure

aks-azure-devops-agents/
├── docker/
│   ├── Dockerfile              # Agent container image (Ubuntu 24.04 LTS)
│   └── start.sh                # Agent startup and registration script
├── helm-chart/
│   └── aks-azure-devops-agents/
│       ├── templates/
│       │   ├── deployment.yaml      # Kubernetes Deployment
│       │   ├── scaledobject.yaml    # KEDA ScaledObject
│       │   ├── secret.yaml          # Azure DevOps PAT secret
│       │   └── trigger-auth.yaml    # KEDA TriggerAuthentication
│       ├── Chart.yaml
│       └── values.yaml
└── pipelines/
    └── azure-pipelines.yaml    # CI/CD pipeline for building and pushing the image

Prerequisites

• Azure Kubernetes Service (AKS) cluster
• Helm v3 installed
• KEDA installed on AKS (KEDA installation guide)
• Azure DevOps account with an agent pool
• Personal Access Token (PAT) with Agent Pools (Read & Manage) permission

Building and Pushing the Docker Image

1. Clone the repository:

   git clone https://github.com/Think-Cube/aks-azure-devops-agents.git
   cd aks-azure-devops-agents/docker

2. Build the Docker image:

   docker build -t <your-acr>.azurecr.io/aks-azure-devops-agents:<tag> .

3. Push the image to Azure Container Registry (ACR):

   az acr login --name <your-acr>
   docker push <your-acr>.azurecr.io/aks-azure-devops-agents:<tag>

Deploying with Helm

1. Copy and edit the values file with your settings:

   cp helm-chart/aks-azure-devops-agents/values.yaml my-values.yaml
   # edit my-values.yaml: set image.repository, azp.url, azp.pool, azp.token, scaledObject.poolID

2. Create Kubernetes namespace:

   kubectl create namespace devops-agents

3. Deploy using Helm:

   helm install azdevops-agents helm-chart/aks-azure-devops-agents \
     --namespace devops-agents \
     --values my-values.yaml

Scaling Behavior

KEDA monitors the number of pending jobs in the Azure DevOps agent pool and automatically scales the number of running agents between minReplicaCount and maxReplicaCount. Scaling is managed via the ScaledObject resource in templates/scaledobject.yaml.

Upgrading

helm upgrade azdevops-agents helm-chart/aks-azure-devops-agents \
  --namespace devops-agents \
  --values my-values.yaml

Uninstalling

helm uninstall azdevops-agents --namespace devops-agents

Configuration

Parameter	Description	Default
replicaCount	Initial number of agent replicas	1
image.repository	Docker image repository	<your-acr>.azurecr.io/aks-azure-devops-agents
image.tag	Image tag	latest
image.pullPolicy	Image pull policy	IfNotPresent
azp.url	Azure DevOps organization URL	https://dev.azure.com/<your-org>
azp.pool	Azure DevOps agent pool name	<your-agent-pool>
azp.token	Base64-encoded Azure DevOps PAT	""
dockerVolumePath	Docker socket path on host	/var/run/docker.sock
scaledObject.minReplicaCount	Minimum running agents	1
scaledObject.maxReplicaCount	Maximum running agents	5
scaledObject.poolID	Azure DevOps agent pool numeric ID	<your-pool-id>
triggerAuthentication.name	KEDA TriggerAuthentication resource name	pipeline-trigger-auth

Finding the pool ID: Go to Organization Settings → Agent pools → select your pool → the poolId query parameter in the URL is the numeric ID.

Troubleshooting

Check agent logs

kubectl logs -f deployment/azdevops-deployment -n devops-agents

Debug KEDA scaling

kubectl get scaledobjects -n devops-agents
kubectl describe scaledobject azure-pipelines-scaledobject -n devops-agents

Agent not registering

• Verify azp.url matches your Azure DevOps organization URL exactly
• Verify the PAT is correctly Base64-encoded and has Agent Pools (Read & Manage) permission
• Check that the pool name in azp.pool matches the pool created in Azure DevOps

License

MIT © Think-Cube

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines.