4 min read

Azure DevOps Multi-Stage Pipelines Approval Strategies

Justin Yoo

There are two ways using Azure DevOps Release Pipelines. We use UI so that we compose each task directly on the screen. On the other hand, we can use YAML pipelines so that all the pipeline stages, jobs and tasks are managed as code. We call the second option as "Multi-Stage Pipelines". As the Multi-Stage Pipelines feature is still in public preview, it doesn't fully offer the same functionalities that the UI pipelines do. The approval process for each release stage is one of the limited features, but it's been available for a couple of months now.

In my previous post, I dealt with YAML pipeline refactoring technics. Throughout this post, I'm going to discuss how each release stage can configure the approval process.

The sample release pipeline code can be found at this GitHub repository.

Prerequisites

It would be nice to have the following tools if you want to read and follow this post:

Multi-Stage Release Pipeline Check-In

The classic UI release pipeline allows developers to put various configurations on each stage. We can configure pre-/post-approvals either automatically or manually. We also can do pre-/post-gated-check-in features. Which one can we do this on our YAML pipelines?

NONE

Unfortunately, we can't do anything like this on the Multi-Stage pipelines. However, the deployment job used in a release stage includes the environment attribute, which allows the manual approval check. Let's have a look at the sample pipeline below. It says the first two release stages have the same environment name of release.

stages:
...
- stage: ReleaseWithoutTemplate
displayName: 'Release without Template'
...
jobs:
- deployment: HostedVs2017
...
environment: release
...
- stage: ReleaseWithStepsTemplate
displayName: 'Release with Steps Template'
...
jobs:
- deployment: HostedVs2017
...
environment: release
...
...
view raw stages.yaml hosted with ❤ by GitHub

In other words, those two different stages share the same environment of release. We can configure the manual approval check on the environment. Let's have a look at the picture below. First of all, click the Environment tab.

Highlighting the Environments tab

In this screen, choose an environment to activate the approval feature. In this example, the picture selects the release environment.

Choosing an environment

Within the release environment, click the three dots button at the right top corner of the screen and select the Checks menu.

Highlighting the Checks option

As it's our first time to add manual approval, we see nothing but the screen below. Click the Create button.

Showing empty screen with the Create button

Select the approvers. If we choose approvers individually, all approvers MUST approve the stage; otherwise, the pipeline cannot proceed. If we want a group or team as an approver, only one of the group/team member would be enough for approval.

Finding approvers in a modal

We've now assigned the approver like below:

Displaying approvers on the Checks screen

Let's rerun the pipeline. The pipeline stops at the stage of Release without Template and waiting for approval. Click the Review button in the middle of the screen.

Highlighting the Review area and Stage waiting

The approver can see the buttons below – to approve or reject. Choose the Approve button to proceed.

Choosing the Approve option

Now the pipeline carries on the stage, stops the next stage of Release with Steps Template and waits for another approval.

Highlighting another Review area and Stage waiting

Give the approval to the pipeline and wait to see. This time, the pipeline doesn't stop at the next stage but keeps moving onto the next stages.

Bypassing the rest stages

The pipeline defined above sets up the environment name of release on the first two stages, and the rest stages use their own environment name. This is the reason why the rest stages are not being interrupted.


You might be able to catch an essential point here. Depending on how we configure the environment on each stage, one environment can dictate all approvals, or multiple environments can take care of their own stages.

Let's have a look at the diagram below. All stages share the release environment and set the QA as the approver. Therefore, every time the stage stops for approval and QA has to approve to move onto the next stage. If we want to skip the approval step at the DEV stage, this single environment approach won't give you that flexibility. If we're going to assign a different approver at the PROD stage, it won't be possible either.

Showing one environment dictating all stages

On the other hand, in the diagram below, each stage has its own environment. We even set a different approver or don't set the approver on each stage. In other words, having a different environment on each stage will give more flexibilities from the approval point of view.

Showing multiple environments looking after their stage only

It may sound more reasonable to have their own environment for each stage. But we also need to consider the management overhead. The more environments we create, the more difficult we manage them. If we have only one environment, we dictate them all in one go.

We can also use a mixture of both approaches. For example, consider group stages into several environments like dev, non-prod and prod and assign stages to each environment like DEV to dev, TEST and UAT to non-prod, and PROD to prod.

As it's still in preview, I'm not sure how it's changing when it becomes GA. But for now, if you need the approval process in the Multi-Stage pipelines, consider those strategies.

If you haven't tried the practice above, why don't you do it now?