Case
How do you deploy Azure Data Factory via a YAML pipeline instead of the
Release pipeline?
 |
| Release ADF pipelines with YAML pipelines |
Solution
In a
previous post we used a YAML pipeline to created an ARM template for ADF. That ARM template is now available as an artifact and ready for deployment. That previous post ended in calling the release part of the pipeline which is in a separate YAML file. This makes it easier to call that same YAML file for test, acceptance and production. Below the last part of the main pipeline:
###################################
# Deploy Test environment
###################################
- stage: DeployTest
displayName: Deploy Test
variables:
- group: ParamsTst
pool:
vmImage: 'windows-latest'
condition: Succeeded()
jobs:
- template: deployADF.yml
parameters:
env: tst
DataFactoryName: $(DataFactoryName)
DataFactoryResourceGroupName: $(DataFactoryResourceGroupName)
DataFactorySubscriptionId: $(DataFactorySubscriptionId)
###################################
# Deploy Acceptance environment
###################################
- stage: DeployAcceptance
displayName: Deploy Acceptance
variables:
- group: ParamsAcc
pool:
vmImage: 'windows-latest'
condition: Succeeded()
jobs:
- template: deployADF.yml
parameters:
env: acc
DataFactoryName: $(DataFactoryName)
DataFactoryResourceGroupName: $(DataFactoryResourceGroupName)
DataFactorySubscriptionId: $(DataFactorySubscriptionId)
###################################
# Deploy Production environment
###################################
- stage: DeployProduction
displayName: Deploy Production
variables:
- group: ParamsPrd
pool:
vmImage: 'windows-latest'
condition: Succeeded()
jobs:
- template: deployADF.yml
parameters:
env: prd
DataFactoryName: $(DataFactoryName)
DataFactoryResourceGroupName: $(DataFactoryResourceGroupName)
DataFactorySubscriptionId: $(DataFactorySubscriptionId)
In this blog we will create the deployADF.yml file mentioned in the YAML code above, but first we need to give the Service Principal (SP), used by the Azure DevOps service connection , access to the target Data Factories otherwise it can't release the ARM template.
1) Access control (IAM)
In this first step we will give the SP access to the target ADF. You have to repeat that for all target factories, but first you have to decide whether you want to give the Service Principal access to specific services (like ADF) or the resource group or even the entire subscription.
In most cases you want to limit the access to the bare minimum to avoid misuse. Since there is no ADF deployment task we are using the more general
AzureResourceManagerTemplateDeployment task. One downside of this task is that you need to give permissions on at least the
resource group. Access to ADF only is not enough and will give you an error: Failed to check the resource group status. Error: {"statusCode":403}.
The next thing to keep in mind is the role to assign. You want to avoid owner to avoid misuse. In this case we need the role Contributor.
- In the Azure portal go to the resource group where your ADF is located
- Click on Access control (IAM) in the left menu
- Click on +Add and then on Add role Assignment
- Search for the appropriate Azure role (this screen recently changed, but you can also still use the classic experience via the link. Click on Contributor and press Next.
- Click on +Select members and search for your SP, click on the account and then press Select
- Optionally add a description and press Next and then Review + assign
 |
| Contributor role in Resource Group for SP |
2) Add additional YAML fileNext step is to add the second YAML file to the repository that does the deployment of ADF. Use the same repository folder as the existing YAML file (in CICD\YAML folder). Splitting up the deployment allows you to reuse the deployment code for test, acceptance and production. Downside is that you have to edit it in the repository instead of under pipelines (but you could also use the YAML extension for Visual Studio Code).
The second YAML consists of 4 parts and the optional
treeview task to check where all your files are located on the agent.
- Parameters and environment
- Treeview
- Stop triggers
- Deploy ADF
- Cleanup and start triggers
A. Parameters and environmentThis YAML file starts with
parameters that will be filled by the main YAML file. As an alternative you could just use the variable group added in the main pipeline because those variables are also available in sub pipelines. There are four string parameters of which only Env has a list of expected/allowed values:
- env (name of the environment: tst, acc or prd)
- DataFactoryName
- DataFactoryResourceGroupName
- DataFactorySubscriptionId
parameters:
- name: env
displayName: Environment
type: string
values:
- dev
- tst
- acc
- prd
- name: DataFactoryName
displayName: Data Factory Name
type: string
- name: DataFactoryResourceGroupName
displayName: Data Factory Resource Group Name
type: string
- name: DataFactorySubscriptionId
displayName: Data Factory Subscription Id
type: string
We also give the job a name and we create an environment. A list of environments can be found under Pipelines - Environments. This is also the place where you can add Approvals and Checks which is not available in the YAML language. The checkout is optional, but very handy when you for example have some custom PowerShell scripts in the repository that you want to execute before, during or after deployment.
jobs:
- deployment: deploymentjob${{ parameters.env }}
displayName: Deployment Job ${{ parameters.env }}
environment: deploy ${{ parameters.env }}
strategy:
runOnce:
deploy:
steps:
- checkout: self
displayName: '1 Retrieve Repository'
clean: true
B. Treeview
This
treeview task is just for debugging. It for example allows you to see where the artifact is located on your agent. This makes it much easier to configure the following steps. When you have finished the pipeline then just delete this task from the steps or
comment it out until you need it again.
###################################
# Show environment and treeview
###################################
- powershell: |
Write-Output "This is the ${{ parameters.env }} environment"
tree "$(Pipeline.Workspace)" /F
displayName: '2 Show environment and treeview Pipeline_Workspace'
C. Stop triggers
Before we can deploy the created ARM template to our ADF we need to make sure nothing is running. Microsoft created a PowerShell
script for this which is included in the generated ARM template folder. It can stop all triggers
 |
| PrePostDeploymentScript.ps1 |
###################################
# Stop triggers
###################################
- task: AzurePowerShell@5
displayName: '3 Stop triggers'
inputs:
azureSubscription: 'sc_adf-devopssp'
pwsh: true
azurePowerShellVersion: LatestVersion
scriptType: filePath
scriptPath: '$(Pipeline.Workspace)\ArmTemplatesArtifact\PrePostDeploymentScript.ps1'
scriptArguments: > # Use this to avoid newline characters in multiline string
-armTemplate '$(Pipeline.Workspace)\ArmTemplatesArtifact\ARMTemplateForFactory.json'
-ResourceGroupName $(DataFactoryResourceGroupName)
-DataFactoryName $(DataFactoryName)
-predeployment $true
-deleteDeployment $false
If you're running your new pipeline and you're getting an error stating that it cannot find your resource group while you're absolutely sure that it exists and that the SP has access to it then please check this blog
post. It shows you how to create a slightly adjusted copy of the script (with an extra parameter) that is stored in the CICD\PowerShel folder.
###################################
# Stop triggers
###################################
- task: AzurePowerShell@5
displayName: '3 Stop triggers'
inputs:
azureSubscription: 'sc_adf-devopssp'
pwsh: true
azurePowerShellVersion: LatestVersion
scriptType: filePath
scriptPath: '$(Pipeline.Workspace)\s\CICD\powershell\PrePostDeploymentADF.ps1'
scriptArguments: > # Use this to avoid newline characters in multiline string
-armTemplate '$(Pipeline.Workspace)\ArmTemplatesArtifact\ARMTemplateForFactory.json'
-ResourceGroupName $(DataFactoryResourceGroupName)
-DataFactoryName $(DataFactoryName)
-predeployment $true
-deleteDeployment $false
-Subscription $(DataFactorySubscriptionId)
 |
| Stopping deployed triggers |
D. Deploy ADF
Now it's finally time for the actual deployment of ADF. As mentioned above we are using the
AzureResourceManagerTemplateDeploymentV3 task. Check the documentation for a description of all parameters.
We will mention one parameter: Deployment Mode.
It is very important to keep this on INCREMENTAL! The complete mode will delete everything in your resource group that is not mentioned in the ARM
template. Since our template only contains ADF you will end up with a nearly empty resource group with only ADF in it. This is a very common mistake. So now you are warned.
###################################
# Deploy ADF Artifact
###################################
- task: AzureResourceManagerTemplateDeployment@3
displayName: '4 Deploy ADF Artifact'
inputs:
deploymentScope: 'Resource Group'
azureResourceManagerConnection: 'sc_adf-devopssp'
subscriptionId: $(DataFactorySubscriptionId)
action: 'Create Or Update Resource Group'
resourceGroupName: $(DataFactoryResourceGroupName)
location: 'West Europe'
templateLocation: 'Linked artifact'
csmFile: '$(Pipeline.Workspace)/ArmTemplatesArtifact/ARMTemplateForFactory.json'
csmParametersFile: '$(Pipeline.Workspace)/ArmTemplatesArtifact/ARMTemplateParametersForFactory.json'
overrideParameters: '-factoryName $(DataFactoryName)'
deploymentMode: 'Incremental'
env:
SYSTEM_ACCESSTOKEN: $(System.AccessToken)
 |
| Deployment of ADF |
E. Cleanup and start triggers
Because we did an incremental deployment all deleted items are still in ADF. Only new and update items have changed. So we have to compare ADF with the template and delete all items that are not in the template. Luckily Microsoft already created a script for this. Same script as the stop trigger script. Just different parameters. And it also enables the triggers.
###################################
# Start triggers and cleanup
###################################
- task: AzurePowerShell@5
displayName: '5 Start triggers and cleanup'
inputs:
azureSubscription: 'sc_adf-devopssp'
pwsh: true
azurePowerShellVersion: LatestVersion
scriptType: filePath
scriptPath: '$(Pipeline.Workspace)\ArmTemplatesArtifact\PrePostDeploymentScript.ps1'
scriptArguments: > # Use this to avoid newline characters in multiline string
-armTemplate $(Pipeline.Workspace)/ArmTemplatesArtifact/ARMTemplateForFactory.json
-ResourceGroupName $(DataFactoryResourceGroupName)
-DataFactoryName $(DataFactoryName)
-predeployment $false
-deleteDeployment $true
Note that the same issue with not finding your Resource Group will occure here if it occured when stopping the triggers. Same solution (different script and extra subscription parameter).
 |
| Start triggers and cleanup |
Now it's time to run the pipeline from start to end by making changes to the Development Data Factory. And in no time all factories are updated.
 |
| The Result |
ConclusionIn this blog post you learned how to use YAML to do the (build and) deployment of ADF in a pipeline. The take away is to use the incremental option and not set it to complete to avoid those shocked looks when viewing your empty resource group.
In a next post we will show how to
overwrite global parameters and
change Linked Services during deployment and show you how to enable or disable certain ADF
triggers depending on the environment. This allows you to have different active triggers in Development, Test, Acceptance and Production without setting them manually after deployment.
Now all YAML parts together:
parameters:
- name: env
displayName: Environment
type: string
values:
- dev
- tst
- acc
- prd
- name: DataFactoryName
displayName: Data Factory Name
type: string
- name: DataFactoryResourceGroupName
displayName: Data Factory Resource Group Name
type: string
- name: DataFactorySubscriptionId
displayName: Data Factory Subscription Id
type: string
jobs:
- deployment: deploymentjob${{ parameters.env }}
displayName: Deployment Job ${{ parameters.env }}
environment: deploy ${{ parameters.env }}
strategy:
runOnce:
deploy:
steps:
- checkout: self
displayName: '1 Retrieve Repository'
clean: true
###################################
# Show environment and treeview
###################################
- powershell: |
Write-Output "This is the ${{ parameters.env }} environment"
tree "$(Pipeline.Workspace)" /F
displayName: '2 Show environment and treeview Pipeline_Workspace'
###################################
# Stop triggers
###################################
- task: AzurePowerShell@5
displayName: '3 Stop triggers'
inputs:
azureSubscription: 'sc_adf-devopssp'
pwsh: true
azurePowerShellVersion: LatestVersion
scriptType: filePath
scriptPath: '$(Pipeline.Workspace)\ArmTemplatesArtifact\PrePostDeploymentScript.ps1'
scriptArguments: > # Use this to avoid newline characters in multiline string
-armTemplate '$(Pipeline.Workspace)\ArmTemplatesArtifact\ARMTemplateForFactory.json'
-ResourceGroupName $(DataFactoryResourceGroupName)
-DataFactoryName $(DataFactoryName)
-predeployment $true
-deleteDeployment $false
###################################
# Deploy ADF Artifact
###################################
- task: AzureResourceManagerTemplateDeployment@3
displayName: '4 Deploy ADF Artifact'
inputs:
deploymentScope: 'Resource Group'
azureResourceManagerConnection: 'sc_adf-devopssp'
subscriptionId: $(DataFactorySubscriptionId)
action: 'Create Or Update Resource Group'
resourceGroupName: $(DataFactoryResourceGroupName)
location: 'West Europe'
templateLocation: 'Linked artifact'
csmFile: '$(Pipeline.Workspace)/ArmTemplatesArtifact/ARMTemplateForFactory.json'
csmParametersFile: '$(Pipeline.Workspace)/ArmTemplatesArtifact/ARMTemplateParametersForFactory.json'
overrideParameters: '-factoryName $(DataFactoryName)'
deploymentMode: 'Incremental'
env:
SYSTEM_ACCESSTOKEN: $(System.AccessToken)
###################################
# Start triggers and cleanup
###################################
- task: AzurePowerShell@5
displayName: '5 Start triggers and cleanup'
inputs:
azureSubscription: 'sc_adf-devopssp'
pwsh: true
azurePowerShellVersion: LatestVersion
scriptType: filePath
scriptPath: '$(Pipeline.Workspace)\ArmTemplatesArtifact\PrePostDeploymentScript.ps1'
scriptArguments: > # Use this to avoid newline characters in multiline string
-armTemplate $(Pipeline.Workspace)/ArmTemplatesArtifact/ARMTemplateForFactory.json
-ResourceGroupName $(DataFactoryResourceGroupName)
-DataFactoryName $(DataFactoryName)
-predeployment $false
-deleteDeployment $true
