Comprehensive guide on the Deployment Jobs strategies in Azure DevOps multi-stage YAML

Comprehensive guide on the Deployment Jobs strategies in Azure DevOps multi-stage YAML

In Azure DevOps Multi-stage YAML pipelines, various deployment strategies can be employed to optimize the deployment process. Below is an exploration of RunOnce, Rolling, and Canary deployment strategies, along with a specific example of Canary Deployment for AKS.

1. RunOnce Deployment Strategy

RunOnce:

This strategy is useful for tasks that should be executed only once, such as initialization, configuration setup, or infrastructure deployment.

xxxxxxxxxx
7
1
stages:
2
  - stage: Setup
3
    jobs:
4
      - job: RunOnceJob
5
        runsOn: self-hosted
6
        steps:
7
          - script: echo "Setting up initial configuration..."

2. Rolling Deployment Strategy

Rolling:

Deploy updates incrementally to minimize risk by moving from lower environments to higher environments, ensuring stability at each stage.

xxxxxxxxxx
18
1
stages:
2
  - stage: Dev
3
    jobs:
4
      - job: DeployDev
5
        steps:
6
          - script: echo "Deploying to Development environment..."
7
  - stage: Staging
8
    dependsOn: Dev
9
    jobs:
10
      - job: DeployStaging
11
        steps:
12
          - script: echo "Deploying to Staging environment..."
13
  - stage: Production
14
    dependsOn: Staging
15
    jobs:
16
      - job: DeployProduction
17
        steps:
18
          - script: echo "Deploying to Production environment..."

3. Canary Deployment Strategy

Canary:

Gradually deploy updates to a small, controlled subset of users or resources to validate changes before full rollout.

xxxxxxxxxx
20
1
stages:
2
  - stage: Canary
3
    jobs:
4
      - job: DeployCanary
5
        steps:
6
          - script: |
7
              kubectl apply -f ./canary-deployment.yaml
8
              kubectl rollout status deployment/myapp -n my-namespace
9
          - hooks:
10
              - approval:
11
                  name: Canary Approval
12
                  condition: eq(variables['Build.Status'], 'Succeeded')
13
  - stage: Production
14
    dependsOn: Canary
15
    jobs:
16
      - job: DeployProduction
17
        steps:
18
          - script: |
19
              kubectl apply -f ./production-deployment.yaml
20
              kubectl rollout status deployment/myapp -n my-namespace

4. Canary Deployment for AKS

Canary for AKS:

Similar to Canary deployment, but specifically for AKS environments, where only a subset of Kubernetes pods are updated and monitored.

xxxxxxxxxx
23
1
stages:
2
  - stage: Canary
3
    jobs:
4
      - job: DeployCanaryAKS
5
        steps:
6
          - script: |
7
              kubectl apply -f ./canary-deployment.yaml
8
              kubectl rollout status deployment/myapp -n my-namespace
9
          - script: |
10
              kubectl get pods -n my-namespace
11
              kubectl logs $(kubectl get pods -n my-namespace | grep myapp-canary | awk '{print $1}') -n my-namespace
12
          - hooks:
13
              - approval:
14
                  name: Canary Approval
15
                  condition: eq(variables['Build.Status'], 'Succeeded')
16
  - stage: Production
17
    dependsOn: Canary
18
    jobs:
19
      - job: DeployProductionAKS
20
        steps:
21
          - script: |
22
              kubectl apply -f ./production-deployment.yaml
23
              kubectl rollout status deployment/myapp -n my-namespace

Summary of Strategies

1. RunOnce: Ideal for setting up configurations, infrastructure, or one-time jobs.

2. Rolling: Deploying updates incrementally, ensuring stability at each step.

3. Canary: Deploying a small portion of resources or users to validate changes before full deployment.

4. Canary for AKS: A specialized version of Canary for Kubernetes, validating and monitoring updates in AKS environments.