Keptn v1 reached EOL December 22, 2023. For more information see https://bit.ly/keptn
This documentation is for an older Keptn release. Please consider the newest one when working with the latest Keptn.

Deployment with Helm

Keptn uses Helm v3 for deploying onboarded services to a Kubernetes cluster. This is currently implemented in the helm-service. Keptn’s helm-service supports the following deployment strategies:

  • Direct deployments
  • Blue-green deployments
  • user-managed deployments (experimental)

The explanation below is based on the provided Helm Chart for the carts microservice, see Charts for details.

Direct deployments

If the deployment strategy of a stage in the shipyard is configured as deploymentstrategy: direct, Helm deploys a release called sockshop-dev-carts as carts in namespace sockshop-dev.

$ kubectl get deployments -n sockshop-dev carts -owide
NAME    READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                 SELECTOR
carts   1/1     1            1           56m   carts        docker.io/keptnexamples/carts:0.11.1   app=carts

When triggering a delivery (with a new artifact), we are updating the values.yaml file in the Helm Charts with the respective image name.

Blue-green deployments

If the deployment strategy of a stage in the shipyard is configured as deploymentstrategy: blue_green_service, Helm creates two deployments within the Kubernetes cluster: (1) the primary and (2) the canary deployment. This can be inspected using the following command:

$ kubectl get deployments -n sockshop-staging carts carts-primary -owide
NAME            READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                 SELECTOR
carts-primary   1/1     1            1           56m   carts        docker.io/keptnexamples/carts:0.11.1   app=carts-primary
carts           0/0     0            0            3m   carts        docker.io/keptnexamples/carts:0.11.2   app=carts

When triggering a delivery (with a new artifact, e.g., 0.11.2), a canary deployment will be modified and scaled up.

NAME            READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                 SELECTOR
carts-primary   1/1     1            1           57m   carts        docker.io/keptnexamples/carts:0.11.1   app=carts-primary
carts           0/1     1            1            1m   carts        docker.io/keptnexamples/carts:0.11.2   app=carts

The primary deployment is always available (and called carts-primary). The canary deployment (called carts) gets scaled up in the case of a new-artifact event (e.g., in this case someone has sent a new-artifact for 0.11.2). Traffic is shifted to the canary release.

Once testing has finished, the primary deployment is upgraded to the new version (0.11.2).

NAME            READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                 SELECTOR
carts-primary   1/1     1            1            3m   carts        docker.io/keptnexamples/carts:0.11.2   app=carts-primary
carts           1/1     1            1            1d   carts        docker.io/keptnexamples/carts:0.11.2   app=carts

After a new pod for the primary deployment has been successfully deployed, traffic is shifted to the primary deployment and the canary deployment is scaled down:

NAME            READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES                                 SELECTOR
carts-primary   1/1     1            1            4m   carts        docker.io/keptnexamples/carts:0.11.2   app=carts-primary
carts           0/0     0            0            1d   carts        docker.io/keptnexamples/carts:0.11.2   app=carts

User Managed deployments (experimental)

⚠️ Note, that this feature is currently marked as experimental. Hence, expect various enhancements and/or changes in future releases of Keptn.

If the deployment strategy of a stage in the shipyard is configured as deploymentstrategy: user_managed, the provided Helm chart will be deployed without any modification and applied as it is.

Assuming that you’ve created a project from a shipyard file containing a stage with deploymentstrategy: user_managed, you need to:

(1) create a service in your stage:

keptn create service <service-name> --project <project-name>

(2) add your desired Helm chart to each stage:

keptn add-resource --project=<project-name> --service=<service-name> --all-stages --resource=<your-helm-chart.tgz> --resourceUri=helm/<service-name>.tgz

(3) upload a file called endpoints.yaml where you can defined the host name under which your deployed service will be available:

keptn add-resource --project=<project-name> --service=<service-name> --all-stages --resource=<path_to_endpoints.yaml> --resourceUri=helm/endpoints.yaml

Note: This step is required, if you will need the data.deployment.deploymentURIsPublic and/or data.deployment.deploymentURIsLocal property of the deployment.finished event sent by the helm service. This is the case, e.g., when the jmeter-service, which performs the test task needs to determine the URL for the service to be tested. The endpoints.yaml file has the following structure:

deploymentURIsLocal:
  - "<my-local-url>" # e.g. http://my-service.sockshop-dev:8080
deploymentURIsPublic:
  - "<my-public-url>" # e.g. http://123.123.123.nip.io:80

(4) send an event to trigger the delivery of your service:

keptn send event --file=./delivery.json

where the content of delivery.json looks something like:

{
  "contenttype": "application/json",
  "data": {
    "project": "<project-name",
    "service": "<service-name>",
    "stage": "<stage-name>"
  },
  "source": "keptn-cli",
  "specversion": "1.0",
  "type": "sh.keptn.event.<stage-name>.delivery.triggered"
}

Limitations:

  • Currently, when using this deployment strategy, you cannot use the keptn trigger delivery command as you would normally do. Hence, you need to send an event using keptn send event like described above.
  • Modifications to the Helm chart via keptn configuration changes are currently not possible.

Clean-up after deleting a project

When executing keptn delete project, Keptn does not clean up existing deployments nor Helm releases. To do so, delete all relevant namespaces:

  • For each stage defined stage within the shipyard of the project, execute kubectl delete namespace <PROJECTNAME>-<STAGENAME>, e.g. for sockshop with stages dev, staging and production:

    kubectl delete namespace sockshop-dev
    kubectl delete namespace sockshop-staging
    kubectl delete namespace sockshop-production