Home / Keptn v1 Docs / Installation / Multi-cluster setup
Keptn Control plane
Keptn Execution plane
To operate Keptn in a multi-cluster setup, you need obviously at least two Kubernetes clusters:
To create a Kubernetes cluster, please follow the instructions here.
The Control Plane of Keptn is responsible for orchestrating your processes for continuous delivery or automated operations.
Before starting the installation, familiarize yourself with the ways of exposing Keptn as explained in Choose access options. Then come back and continue installing the Keptn control plane.
To install the control plane, execute
helm install with the option you chose for exposing Keptn:
helm install keptn keptn/keptn -n keptn --version=$KeptnVersion --create-namespace --set=apiGatewayNginx.type=[LoadBalancer, NodePort, ClusterIP]
Before continuing, please retrieve:
KEPTN_ENDPOINT: Follow the instructions for the option you chose for exposing Keptn in the Choose access option section.
KEPTN_API_TOKEN: Follow the instructions in Authenticate Keptn CLI and Bridge.
In this release of Keptn, the execution plane services for deployment (
helm-service) and testing (
jmeter-service) can be installed via Helm Charts.
Please find the Helm Charts here:
helm-service: GitHub Release for 0.19.3 at Assets >
jmeter-service: GitHub Release for 0.19.3 at Assets >
values.yaml from the release branch, e.g., for the jmeter-service:
values.yaml to connect the services to the Keptn control plane, identified by its endpoint and API token. Therefore, set the values (1) - (5):
remoteControlPlane: enabled: true # < (1) set to true api: protocol: "http" # < (2) set protocol: http or https hostname: "" # < (3) set Keptn endpoint (without /api) apiValidateTls: true # < (4 - optional) option to skip TLS verification token: "" # < (5) set Keptn API token
If your cluster includes multiple execution planes that run the same integration service, you must configure a unique name for each execution plane. By default, Keptn uses the execution plane service name and version to identify the execution plane. Multiple execution planes that run the same integration service thus have the same identifier, so only one instance of the integration service is displayed on the integration page in the Bridge and Keptn assigns the same set of event subscriptions to them all unless you assign a different name to each remote execution plane service.
You can assign a unique name label and distributor service name for the execution plane in either of the following ways:
K8S_DEPLOYMENT_NAME environment variable on each execution plane to a unique name.
See the distributor reference page
for more information about this and other environment variables that configure the distributor.
Some integrations (such as
can configure the
based on the value of the
app.kubernetes.io/name Kubernetes label.
For these integrations, you can edit the
values.yaml on each execution plane
and set a unique value for the
Depending on your setup of the multi-cluster environment and the approach you modeled your staging process, one stage can be for example on a separate cluster. Let’s assume the following setup:
hardening- on Cluster-A
production- on Cluster-B
To properly configure the execution plane services that run, for example, on Cluster-A, the distributor in the
values.yaml needs to be configured:
distributor: projectFilter: "" # set the project, e.g., "sockshop" to get events for this project. stageFilter: "hardening" # set the stage, e.g., "hardening" to get events for the stage. serviceFilter: "" # set the service, e.g., "carts" to get events for the service.
serviceFilter allow a comma-separated list of values.
Deploy the execution plane service (e.g., jmeter-service) from release assets with your
values.yaml and by using
helm install jmeter-service https://github.com/keptn/keptn/releases/download/0.19.3/jmeter-service-0.19.3.tgz -n keptn-exec --create-namespace --values=values.yaml
Test connection to Keptn control plane using:
helm test jmeter-service -n keptn-exec
Pod jmeter-service-test-api-connection pending Pod jmeter-service-test-api-connection succeeded NAME: jmeter-service LAST DEPLOYED: Thu Feb 25 15:55:24 2021 NAMESPACE: keptn-exec STATUS: deployed REVISION: 1 TEST SUITE: jmeter-service-test-api-connection Last Started: Thu Feb 25 15:55:40 2021 Last Completed: Thu Feb 25 15:55:42 2021 Phase: Succeeded
To uninstall an execution plane service – in this case, the
helm uninstall jmeter-service -n keptn-exec
See the configuration parameters of the supported execution plane services:
helm-service: Helm Chart values
jmeter-service: Helm Chart values
The important ones that are used in the above example are:
||Sets the stage to which this service||
||Sets the service to which this service||
||Sets the project to which this service||
||Enables remote execution plane mode||
||Used protocol (http, https)||
||Hostname of the control plane cluster (and port)||
||Defines if the control plane certificate should be validated||
||Keptn API token||
||Sets a unique name for this execution plane||
If you see in the Keptn Bridge that an event was triggered but no service was reacting upon this trigger, test the connectivity from the execution plane service to the control plane. (as mentioned above) The Helm Charts for the
jmeter-service have a built in sanity check that validates whether the connection to the control plane can be established.
Test (sanity check):
Connect you to the cluster where the execution plane is running
For example, to test
jmeter-service that is running in the
keptn-exec namespace, execute:
helm test jmeter-service -n keptn-exec
The expected outcome is:
Pod jmeter-service-test-api-connection pending Pod jmeter-service-test-api-connection succeeded NAME: jmeter-service LAST DEPLOYED: Thu Feb 25 15:55:24 2022 NAMESPACE: keptn-exec STATUS: deployed REVISION: 1 TEST SUITE: jmeter-service-test-api-connection Last Started: Thu Feb 25 15:55:40 2022 Last Completed: Thu Feb 25 15:55:42 2022 Phase: Succeeded
remoteControlPlane: enabled: true # < (1) set to true api: protocol: "http" # < (2) set protocol: http or https hostname: "" # < (3) set Keptn hostname (without /api) apiValidateTls: true # < (4 - optional) option to skip TLS verification token: "" # < (5) set Keptn API token
The same integration service can be deployed to multiple execution planes in your Keptn installation.
For example, you might deploy the
to the execution plane that runs the
dev stage as well as the execution plane that runs the
However, you must explicitly configure each execution plane name as described above
so that Keptn can differentiate the two instances of the integration.
If you do not correctly configure the execution plane names, you will see two issues:
helm-service) is displayed on the integration page in the Keptn Bridge.
devstage also receives events intended for the
To correct this situation, follow the instructions above to set a unique name label and distributor service name for each execution plane that runs this integration service.