Modern CI/CD with GKE: Apply the developer workflow

Last reviewed 2023-09-11 UTC

This tutorial shows how to onboard a new application, develop a feature for the application, and deploy the application to production using modern continuous integration/continuous delivery (CI/CD) techniques with Google Kubernetes Engine (GKE).

This document is part of a series:

In this tutorial, you use tools such as Skaffold, kustomize, Artifact Registry, Config Sync, Cloud Build and Cloud Deploy to develop, build and deploy your application.

This document is intended for enterprise architects and application developers, as well as IT security, DevOps, and Site Reliability Engineering (SRE) teams. Some experience with automated deployment tools and processes is useful for understanding the concepts in this document.

Architecture

In this tutorial, you onboard a new application. Then, you develop a new feature and deploy the application in dev, staging, and production environments. The reference architecture contains the infrastructure and tools needed to onboard and release a new application with the workflow shown in the following diagram:

The development loop spans code repositories, cloudbuild and cloud deploy pipelines.

Starting at the code repository for CI, the workflow includes the following steps:

  1. You share your application source code through your application repositories.

  2. When you commit and push the code to the application repository, it automatically triggers a CI pipeline in Cloud Build. The CI process creates and pushes a container image to Artifact Registry.

  3. The CI process also creates a CD release for the application in Cloud Deploy.

  4. The CD release generates fully rendered kubernetes manifests for dev using skaffold and deploys them in the dev GKE cluster.

  5. The CD release is then promoted from dev to a staging target, that generates fully rendered staging manifests and deploys them in the staging GKE cluster.

  6. Then, The CD release is promoted from staging to production, that generates fully rendered production manifests and deploys them in production GKE clusters.

For more information about the tools and infrastructure used in this workflow, see Modern CI/CD with GKE: Build a CI/CD system.

Objectives

  • Onboard a new application.

  • Deploy the application in the dev environment.

  • Develop a new feature and deploy it in the dev environment.

  • Promote the new feature to staging and then release it into production.

  • Test the resiliency of the application.

Costs

In this document, you use the following billable components of Google Cloud:

To generate a cost estimate based on your projected usage, use the pricing calculator. New Google Cloud users might be eligible for a free trial.

When you finish the tasks that are described in this document, you can avoid continued billing by deleting the resources that you created. For more information, see Clean up.

Before you begin

Prepare your environment

  1. If you're continuing directly from Modern CI/CD with GKE: Build a CI/CD system, go to the next section. However, if you have a new session, or your session expired, open Cloud Shell and set the project where you installed the reference architecture infrastructure:

    gcloud config set core/project PROJECT_ID

    Replace PROJECT_ID with your Google Cloud project ID.

Onboard a new application

The reference architecture contains an application factory. This factory is a collection of a git repository named application-factory-repo and the following Cloud Build triggers:

  • create-app
  • tf-plan
  • tf-apply
  • create-team

You use the application factory to onboard a new application from starter repositories. Application onboarding consists of the following steps:

  1. Create the application definition: You create the application definition in a Terraform file and store it in application-factory-repo which acts as application catalog.

  2. Create the application infrastructure: You run Terraform on the application definition file to create the application infrastructure. Application infrastructure consists of the following:

    1. A Landing zone for the new application includes defining the namespace, service account, and base policies in the acm-gke-infrastructure-repo repository. The landing zone is only created in a dev GKE cluster while onboarding a new application. This is done to unblock the developers so that they can use the dev environment and start iterating on it. The landing zone in the staging and production clusters are created with the GitOps approach. This approach is demonstrated later in this document when you are ready to promote the release in those clusters.

    2. The Infrastructure repository from infrastructure starter repository that hosts the code to create the CI pipeline in Cloud Build, the CD pipeline in Cloud Deploy, and the Artifact Registry repository for storing artifacts.

    3. An infrastructure Cloud Build trigger that takes the code in the infrastructure repository and creates the resources based on their definition.

    4. An application repository from the application starter repository that hosts the application's source code.

  3. Create application's CI/CD resources: You use the application infrastructure to create CI/CD resources for the application.

Create application definition:

Run the create-app trigger to generate an application definition file in application-factory-repo. The definition file contains the declarative definition of the resources required to create an application.

  1. In the Google Cloud console, go to the Cloud Build page:

    Go to the Cloud Build page

  2. Click create-app trigger.

  3. Click SHOW URL PREVIEW to display the URL required to invoke the webhook.

  4. In Cloud Shell, invoke the trigger by making a curl request on the URL obtained from the previous step and passing the parameters as a payload to it.

    curl "WEBHOOK_URL" -d '{"message": {"app": "sample","runtime": "python","trigger_type": "webhook","github_team": ""}}'

    In the previous code sample:

    • Replace WEBHOOK_URL with the URL obtained from the trigger.

    • "app": "sample" specifies the name of the application.

    • "runtime": "python" tells the application factory to use the python template to create application repositories.

    • "trigger_type": "webhook" specifies the type of CI/CD pipelines for the application.

    • "github_team": "" is a team in GitHub that's going to be associated with the repositories that are created for the application. Since, you haven't created a GitHub team yet, pass it as an empty string.

  5. Check the pipeline for create-app trigger:

    Go to the Cloud Build History page.

    There is a new pipeline for create-app trigger. When that has completed, the application definition is created in application-factory-repo.

  6. Review the application definition file:

    1. In a web browser, go to GitHub and sign in to your account.

    2. Click the picture icon and click Your organizations. Choose your organization.

    3. Click the repository application-factory-repo, go to the folder apps/python and open the new file named sample.tf created by the create-app trigger. Inspect the file, this file contains Terraform code to create a new application.

Create application infrastructure:

Now that you've created the application definition, you run the trigger tf-apply to create the application infrastructure.

  1. In Google Cloud console:

    Go to the Cloud Build page .

    Click tf-apply trigger.

  2. Click "SHOW URL PREVIEW" to display the URL required to invoke the webhook.

  3. Invoke the trigger:

    curl "WEBHOOK_URL" -d '{}'

    In the previous code sample:

    • Replace WEBHOOK_URL with the URL obtained from the trigger.
  4. Check the pipeline for tf-apply trigger:

    Go to the Cloud Build History page.

    There is a new pipeline for tf-apply trigger. Wait for it to complete.

This trigger creates the application infrastructure.

Review the application infrastructure:

Review the various components of the application infrastructure.

Landing zone

  1. Go to Cloud Shell and set the project.

    gcloud config set core/project PROJECT_ID

    Replace PROJECT_ID with your Google Cloud project ID.

  2. Get credentials to the development GKE cluster.

    gcloud container clusters get-credentials gke-dev-us-central1 --zone us-central1-a
    
  3. Check the namespace of the application. The namespace is named after the application, sample.

    kubectl get namespaces sample
    

    The output resembles the following:

    NAME     STATUS   AGE
    sample   Active   15m
    

  4. Check the service account in the namespace.

    kubectl get serviceaccounts -n sample
    

    There is a service account in addition to the default. The output resembles the following:

    NAME         SECRETS   AGE
    default      0         15m
    sample-ksa   0         15m
    

Infrastructure repository

In a web browser, go to GitHub and sign in to your account. Click the picture icon. Then click Your organizations. Choose you organization and click the sample-infra repository.

This repository has four branches: cicd-trigger, dev, staging and prod. It also contains four folders cicd-trigger, dev, staging and prod. The default branch is cicd-trigger and you can push the code to it while other branches have protection rules so, you can't push code directly to those branches. In order to push the code to those branches, you need to create a pull request. The cicd-trigger folder has code to create CI/CD resources for the application while dev, staging and prod folders have code to create infrastructure for different environments of the application.

Infrastructure trigger

  • In Google Cloud console:

    Go to the Cloud Build page

    There is a new trigger named deploy-infra-sample.

  • This trigger is connected to the repository sample-infra such that when a code push happens to this repository, the trigger is invoked and identifies the branch where the push happened and goes to the corresponding folder on that branch and runs Terraform there. For example, if the code is pushed to cicd-trigger branch, the trigger runs Terraform on cicd-trigger folder of cicd-trigger branch. Similarly, when a push happens to the dev branch, the trigger runs Terraform on dev folder of the dev branch and so forth.

Application repository

  • Go to GitHub and pull up the repositories under your organization. There is a new repository with the name sample. This repository hosts the source code and steps to build containers in Dockerfile, kustomize configs that describe needed configurations of the application and skaffold.yaml that defines the deployment steps to be used by Cloud Deploy for CD.

Create application CI/CD resources

Now that you've created the skeleton of the application, run the trigger deploy-infra-sample to create its CI/CD resources. You can either invoke the trigger manually using its webhook URL or by making a commit to the git repository sample-infra.

  1. To invoke the Cloud Build trigger, add a new line in a file in the repository. Then, push the changes:

    1. If you've never used Git in Cloud Shell, configure Git with your name and email address. Git uses this information to identify you as the author of the commits that you create in Cloud Shell:

      git config --global user.email "GITHUB_EMAIL_ADDRESS"
      git config --global user.name "GITHUB_USERNAME"

      Replace the following:

      • GITHUB_EMAIL_ADDRESS: the email address associated with your GitHub account
      • GITHUB_USERNAME: the username associated with your GitHub account
    2. Clone the git repository sample-infra:

      git clone https://github.com/GITHUB_ORG/sample-infra
      
      cd sample-infra

      Replace the following:

      • GITHUB_ORG with your GitHub organization.

      The default branch cicd-trigger is checked out.

    3. Add a new line to the file env/cicd-trigger/main.tf, commit the change and push.

        echo "" >> env/cicd-trigger/main.tf
      
    4. Commit and push the changes:

      git add .
      git commit -m "A dummy commit to invoke the infrastrucutre trigger"
      git push
      cd ..
      

      As soon as the changes are pushed, the Cloud Deploy trigger deploy-infra-sample is started.

  2. Monitor the status of the trigger:

    Go to Cloud Build history page to view the pipeline and wait for it to complete.

Review application CICD resources

Review the various CI/CD resources created for the application.

  1. In the Google Cloud console:

    Go to the Cloud Build page and view the deploy-app-sample trigger.

    This is the CI pipeline trigger. It's connected to the application code repository sample. The trigger gets invoked when a push is made to the application repository and performs the build steps as defined in the trigger configuration. To view the steps that the trigger performs when invoked, click the trigger name and then click OPEN EDITOR button.

  2. Go to the Artifact Registry page and view the new repository with the name sample.

    This Artifact Repository stores the application's artifacts.

  3. Go to the Cloud Deploy pipeline page and view the pipeline with the name sample. This is the continuous deployment pipeline that deploys the application on the GKE clusters.

Deploy the application in the dev environment

The trigger deploy-app-sample is connected to the application repository named sample. You can either invoke the trigger manually, using the webhook URL, or through a push to the application repository.

  1. Add a new line in a file in the repository sample and push the changes to invoke the Cloud Build trigger:

    1. Clone the git repository sample:

      In Cloud Shell:

      git clone https://github.com/GITHUB_ORG/sample
      
      cd sample

      Replace GITHUB_ORG with your GitHub prganization.

    2. Add a new line to the skaffold.yaml file.

        echo "" >> skaffold.yaml
      
    3. Commit and push the changes:

      git add .
      git commit -m "A dummy commit to invoke CI/CD trigger"
      git push
      
    4. As soon as the changes are pushed, the Cloud Deploy trigger deploy-app-sample is started.

  2. Monitor the status of the trigger:

    Go to Cloud Build history page to view the pipeline and wait for it to complete.

    The trigger executes the steps defined in its configuration. The first step is to build a docker image from the application code in the repository sample. The last step is to kick-off the Cloud Deploy pipeline that deploys the application to the dev GKE cluster.

  3. Check the deployment in dev cluster:

    Go to Cloud Deploy pipeline page.

    Click sample pipeline, the deployment to the dev GKE cluster has been started. Wait for it to complete.

Verify that the application has been deployed successfully :

  1. Get credentials to the development cluster.

    gcloud container clusters get-credentials gke-dev-us-central1 --zone us-central1-a
    
  2. Tunnel into GKE cluster.

    gcloud container clusters get-credentials gke-dev-us-central1 --zone us-central1-a && kubectl port-forward --namespace sample $(kubectl get pod --namespace sample --selector="deploy.cloud.google.com/delivery-pipeline-id=sample" --output jsonpath='{.items[0].metadata.name}') 8080:8080
    
  3. On the Cloud Shell toolbar, click

    Web Preview, and then click Preview on port 8080: Cloud Shell toolbar commands.

    The output is the following:

    Hello World!
    
  4. In Cloud Shell, press CTRL+C to end the port forward.

Add a new feature to the application

When you develop a new feature, you need to quickly deploy your changes into development environment in order to test and iterate on them. In this tutorial, you use make changes in application code repository and deploy them to development environment.

  1. In Cloud Shell, change directory to already cloned sample repository:

  2. Update the application to output a different message:

      sed -i "s/Hello World/My new feature/g" main.py
    
  3. Commit and push the changes:

    git add .
    git commit -m "Changed the message"
    git push
    

    As soon as the code is pushed to the GitHub repository, the webhook trigger deploy-app-sample gets kicked-off.

  4. Monitor the status of the trigger on Cloud Build history page and wait for it to complete.

  5. Go to Cloud Deploy pipeline page

    Click sample pipeline, the deployment to the dev GKE cluster has been started. Wait for it to complete.

Verify that the application has been deployed successfully :

  1. Get credentials to the development cluster if you opened a new Cloud Shell:

    gcloud container clusters get-credentials gke-dev-us-central1 --zone us-central1-a
    
  2. Tunnel into GKE cluster:

    gcloud container clusters get-credentials gke-dev-us-central1 --zone us-central1-a && kubectl port-forward --namespace sample $(kubectl get pod --namespace sample --selector="deploy.cloud.google.com/delivery-pipeline-id=sample" --output jsonpath='{.items[0].metadata.name}') 8080:8080
    
  3. On the Cloud Shell toolbar, click

    Web Preview, and then click Preview on port 8080:

    Cloud Shell toolbar commands.

    The output is the following:

    My new feature!
    
  4. In Cloud Shell, press CTRL+C to end the port forward.

Promote your change to the staging and productions clusters

Before promoting the application to staging and production environments, you need to create the landing zone for the application in the GKE clusters for those environments. When you onboarded the application, the landing zone for dev was automatically created in dev GKE cluster by adding code to acm-gke-infrastructure-repo in the dev branch.

Create landing zone in staging and prod GKE clusters

  1. Create landing zone in staging GKE cluster: You need to create a pull request from dev to staging branch in acm-gke-infrastructure-repo and merge it.

    1. Go to GitHub and traverse to the repository acm-gke-infrastructure-repo. Click Pull requests and then New pull request button. In the Base menu, choose staging and in the Compare menu, choose dev. Click the Create pull request button.

    2. Typically, someone who has access to the repository reviews the changes and then merge the PR to make sure only the intended changes are being promoted to the staging environment. To let the individuals try out the reference architecture, the branch protection rules have been relaxed so that the repository administrator are able to bypass the review and merge the PR. If you are an administrator on the repository, merge the pull request. Otherwise, get the administrator to merge it.

    Config Sync syncs the changes that arrive on staging branch of the repository acm-gke-infrastructure-repo with the staging GKE cluster that results in landing zone creation for the application on the staging GKE cluster.

  2. Create landing zone in production GKE clusters: You need to create a pull request from staging to prod branch and merge it.

    1. Click Pull requests and then New pull request button. In the Base menu, choose prod and in the Compare menu, choose staging. Click the Create pull request button.

    2. If you are an administrator on the repository, merge the pull request. Otherwise, get the administrator to merge it.

    Config Sync syncs the changes that arrive on prod branch of the repository acm-gke-infrastructure-repo with the production GKE clusters that results in landing zone creation for the application on production GKE clusters.

Promote the changes from dev to staging

Now that you've created the landing zone for the application in staging and prod GKE clusters, promote the application from dev to staging environment.

  1. Find the latest release name and save it as an environment variable:

      export RELEASE=$(gcloud deploy targets describe dev --region=us-central1 --format="json" | jq -r '."Active Pipeline"[0]."projects/PROJECT_ID/locations/us-central1/deliveryPipelines/sample"."Latest release"' | awk -F '/' '{print $NF}')
     

    Replace PROJECT_ID with your Google Cloud project ID.

    Verify the environment variable has been set:

      echo $RELEASE
      

  2. In Cloud Shell, run the following command to trigger promotion of the release from dev to staging environment:

     gcloud deploy releases promote --release=$RELEASE --delivery-pipeline=sample  --region=us-central1 --to-target=staging --quiet
     

  3. Check staging deployment:

    Go to Cloud Deploy pipeline page

    Click sample pipeline, the deployment to the staging GKE cluster has been started. Wait for it to complete.

  4. Verify the staging deployment has happened successfully:

    1. Get credentials to the staging cluster:

      gcloud container clusters get-credentials gke-staging-us-central1 --zone us-central1-a
      
    2. Tunnel into GKE cluster:

      gcloud container clusters get-credentials gke-staging-us-central1 --zone us-central1-a && kubectl port-forward --namespace sample $(kubectl get pod --namespace sample --selector="deploy.cloud.google.com/delivery-pipeline-id=sample" --output jsonpath='{.items[0].metadata.name}') 8080:8080
      
    3. On the Cloud Shell toolbar, click

      Web Preview, and then click Preview on port 8080:

      Cloud Shell toolbar commands.

      The output is the following:

      My new feature!
      
    4. In Cloud Shell, press CTRL+C to end the port forward.

Promote the changes from staging to prod

Now, promote the release from staging to prod. You have two production clusters and Cloud Deploy has a target for each of them named prod1 and prod2 respectively.

  1. In Cloud Shell, run the following command to trigger promotion of the release from staging to prod1 cluster:

     gcloud deploy releases promote --release=$RELEASE --delivery-pipeline=sample  --region=us-central1 --to-target=prod1 --quiet
     

  2. Release to prod clusters require approval so the rollout waits until you approve it. To view it:

    Go to Cloud Deploy pipeline page

    Click sample pipeline. The rollout to prod1 requires approval and clouddeploy.approver role is needed to approve the rollout. Since, you are the owner of the project, you have access to approve the release.

  3. Approve the release to prod1:

    Run the following command to fetch the name of the rollout pending approval and save it in an environment variable:

     export ROLLOUT=$(gcloud deploy targets describe prod1 --region=us-central1 --format="json" | jq -r '."Pending Approvals"[]' | awk -F '/' '{print $NF}')
     

    Approve the release:

     gcloud deploy rollouts approve $ROLLOUT --delivery-pipeline=sample --region=us-central1 --release=$RELEASE --quiet
     

  4. After the approval is given, prod1 release starts. Monitor the progress on Cloud Deploy pipeline page.

  5. After prod1 deployment is completed, kick-off prod2 release.

     gcloud deploy releases promote --release=$RELEASE --delivery-pipeline=sample  --region=us-central1 --to-target=prod2 --quiet
     

  6. Release to prod2 also requires approval. Approve the release to prod2 cluster:

    Run the following command to fetch the name of the rollout pending approval and save it in an environment variable:

     export ROLLOUT=$(gcloud deploy targets describe prod2 --region=us-central1 --format="json" | jq -r '."Pending Approvals"[]' | awk -F '/' '{print $NF}')
     

    Approve the release:

     gcloud deploy rollouts approve $ROLLOUT --delivery-pipeline=sample --region=us-central1 --release=$RELEASE --quiet
     

  7. After the approval is given, prod2 release starts. Monitor the progress on Cloud Deploy pipeline page.

  8. Verify the deployment in production cluster was successful after the Cloud Deploy pipelines in prod1 and prod2 are completed.

    1. There is Multi Cluster Ingress created in production clusters and you use a load balancer to access the prod application. This Multi Cluster Ingress configurations are created using YAML files k8s/prod/mci.yaml and k8s/prod/mcs.yaml in the repository sample. When you send request to the IP address of the load balancer, Multi Cluster Ingress forwards the request to one of the two instances of the application running in two different GKE clusters.

    2. List the forwarding rule associated to the load balancer to find the IP address.

      gcloud compute forwarding-rules list
    3. The output resembles the following:

      NAME: mci-qqxs9x-fw-sample-sample-ingress
      REGION:
      IP_ADDRESS: 34.36.123.118
      IP_PROTOCOL: TCP
      TARGET: mci-qqxs9x-sample-sample-ingress
      

    4. Open a web browser and enter the following in the URL:

      http://IP_ADDRESS:80

      Replace IP_ADDRESS with IP address of the load balancer.

      The output is the following:

      My new feature!
      

      This confirms the application is deployed as expected in production clusters.

Test resiliency of the application

In this section you test the resiliency of the application running in production by re-starting one of the two production GKE clusters' nodes without impacting the application.

The application in production uses multi-cluster ingress and is accessible through a load balancer IP. When the application is accessed through that IP, multi-cluster ingress routes it to one of the two instances of the application running on two different GKE clusters. When one of the GKE clusters is unhealthy and the application instance running on it can't be reached, the multi-cluster ingress keeps sending the traffic to the healthy instance of the application running on the other GKE cluster. This makes the cluster outage invisible to the end user and the application continuously serves the requests.

To test the resiliency:

  1. Find the node pool of the production GKE clusters running in us-west1.

     gcloud container clusters describe gke-prod-us-west1 --zone=us-west1-a --format=json | jq ".nodePools[0].instanceGroupUrls[]" | tr '"' ' ' |  awk -F '/' '{for(i=NF-2; i<=NF; i=i+2) printf ("%s ",$i); print  ""}'

  2. The output resembles the following:

    us-west1-b gke-gke-prod-us-west1-node-pool-01-6ad4e1ed-grp
    us-west1-c gke-gke-prod-us-west1-node-pool-01-98407373-grp
    

    The output has two columns, first column is the zone and the second column is the name of the instance group associated with the node pool of the production GKE cluster in us-west1 region.

  3. Restart the instance-group corresponding to the node pools:

     gcloud compute instance-groups managed rolling-action restart INSTANCE_GROUP_1  --zone=ZONE_1 --max-unavailable=100%
    
     gcloud compute instance-groups managed rolling-action restart INSTANCE_GROUP_2  --zone=ZONE_2 --max-unavailable=100%

    Replace INSTANCE_GROUP_1 with the name of the first instance group.

    Replace ZONE_1 with the zone of the first instance group.

    Replace INSTANCE_GROUP_2 with the name of the second instance group.

    Replace ZONE_2 with the zone of the second instance group.

  4. Check the status of the instance group.

    Go to Instance groups page

    The two instance groups are being rebooted while the other groups have a green tickmark against them.

    Status of the instance groups.

  5. Open a web browser and enter the following in the URL:

    http://IP_ADDRESS:80

    Replace IP_ADDRESS with IP address of the load balancer.

    Even when one of the two GKE clusters is down, the application is available and the output is the following:

     My new feature!
     

    This shows that your application is resilient and highly available.

Manage the application

When you created this application from the application factory, you got separate git repos, infrastrucutre and CI/CD pipelines for the application. You used these resources to deploy the application and add a new feature. In order to manage the application further, you only need to interact with these git repos and the pipeline without needing to update the application factory. You can customize the application's pipelines and git repositories based on your requirements. As an application owner, you can define who gets access to the pipelines and git repos of your application to manage it.

Clean up

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial.

Delete the project

  1. In the Google Cloud console, go to the Manage resources page.

    Go to Manage resources

  2. In the project list, select the project that you want to delete, and then click Delete.
  3. In the dialog, type the project ID, and then click Shut down to delete the project.

What's next