Kubernetes phased release

Background information

• Kubernetes uses a Service to expose applications within a cluster.

• A Service connects to the Pod instances of a Deployment using labels. The Service then routes traffic to the associated Pod instances.

• Assume a Service named serviceA is associated with an application app-v1 that has more than one replica (replicas > 1). When a phased release begins, a new version, app-v2, is created and inherits all the labels from app-v1. At this point, the Service randomly distributes traffic to v1 and v2.

• The phased release process updates the application by scaling up app-v2 and scaling down app-v1. If the release is paused, you can either resume it or roll it back. A rollback removes app-v2 and restores app-v1 to its original replica count.

Prerequisites

• The Deployment must not be associated with an HPA. Otherwise, the old version cannot be scaled down, which may cause the release to time out and fail.

• The labels in the Deployment's Pod template must match the selector of the associated Service. For example:

  ---
  apiVersion: apps/v1
  kind: Deployment
  metadata:
    labels:
      run: spring-boot-sample
    name: spring-boot-sample
  spec:
    replicas: 2
    selector:
      matchLabels:
        run: spring-boot-sample
    template:
      metadata:
        labels:
          run: spring-boot-sample
      spec:
        containers:
        - image: ${IMAGE}
          name: app
  ---
  apiVersion: v1
  kind: Service
  metadata:
    name: spring-boot-service
    labels:
      test: test
      labele2: label
  spec:
    ports:
      - name: http
        port: 8080
        protocol: TCP
        targetPort: 8080
      - name: https
        port: 443
        protocol: TCP
        targetPort: 443
    selector:
      run: spring-boot-sample
    sessionAffinity: None
    type: ClusterIP

Add a phased release task

In the Flow console, select the target pipeline and add a "Kubernetes Phased Release" task. On the workflow configuration tab, add a new stage. In the task panel, select the Deployment category and click the Kubernetes Phased Release card.

Configure the release task

In this example, for Cluster connection, select testing. For kubectl version, select v1.27.9. For Namespace, enter default. For Service, enter node-expressjs-sample. For Associated workload type, select Deployment. For Phased release strategy, select Pause After First Batch. For Number of batches, enter 2. For Container name, enter node-expressjs-sample. For Image, select Build image and push to ACR Personal Edition.public image address.

Execute the phased release

1. Save and run the pipeline. The pipeline automatically pauses at the Kubernetes Phased Release stage after the first batch is complete. The task card displays a paused status. Click Resume Release to continue with the remaining batches, or click Roll Back Release to revert to the previous version.

2. After you complete manual verification and resume the release, the pipeline deploys the subsequent batches. After all batches are complete, the pipeline status changes to Succeeded. The Build stage (Build image and push to ACR Personal Edition) and the Kubernetes Phased Release stage both show a green check mark. A Release Succeeded message appears below the deployment target.