open-cluster-management.io/placement

Placement

With Placement, you can select a set of ManagedClusters from the ManagedClusterSets bound to the placement namespace.

Community, discussion, contribution, and support

Check the CONTRIBUTING Doc for how to contribute to the repo.

Getting Started

Prerequisites

You have at least one running kubernetes cluster;

Clone this repo

git clone https://github.com/open-cluster-management-io/placement.git
cd placement

Deploy the placement controller

Set environment variables.

export KUBECONFIG=</path/to/kubeconfig>

Build the docker image to run the placement controller.

go install github.com/openshift/imagebuilder/cmd/imagebuilder@v1.2.1
make images
export IMAGE_NAME=<placement_image_name> # export IMAGE_NAME=quay.io/open-cluster-management/placement:latest

If your are using kind, load image into the kind cluster.

kind load docker-image <placement_image_name> # kind load docker-image quay.io/open-cluster-management/placement:latest

And then deploy placement manager on the cluster.

make deploy-hub

What is next

After a successful deployment, check on the cluster and see the placement controller has been deployed.

kubectl -n open-cluster-management-hub get pods
NAME                                                  READY   STATUS    RESTARTS   AGE
cluster-manager-placement-controller-cf9bbd6c-x9dnd   1/1     Running   0          2m16s

Here is an example.

Create a ManagedClusterSet.

cat <<EOF | kubectl apply -f -
apiVersion: cluster.open-cluster-management.io/v1beta2
kind: ManagedClusterSet
metadata:
  name: clusterset1
EOF

Create a ManagedCluster and assign it to clusterset clusterset1.

cat <<EOF | kubectl apply -f -
apiVersion: cluster.open-cluster-management.io/v1
kind: ManagedCluster
metadata:
  name: cluster1
  labels:
    cluster.open-cluster-management.io/clusterset: clusterset1
    vendor: OpenShift
spec:
  hubAcceptsClient: true
EOF

Create a ManagedClusterSetBinding to bind the ManagedClusterSet to the default namespace.

cat <<EOF | kubectl apply -f -
apiVersion: cluster.open-cluster-management.io/v1beta2
kind: ManagedClusterSetBinding
metadata:
  name: clusterset1
  namespace: default
spec:
  clusterSet: clusterset1
EOF

Now create a Placement:

cat <<EOF | kubectl apply -f -
apiVersion: cluster.open-cluster-management.io/v1beta1
kind: Placement
metadata:
  name: placement1
  namespace: default
spec:
  predicates:
    - requiredClusterSelector:
        labelSelector:
          matchLabels:
            vendor: OpenShift
EOF

Check the PlacementDecision created for this placement. It contains all selected clusters in status.

kubectl get placementdecisions
NAME                    AGE
placement1-decision-1   2m27s

kubectl describe placementdecisions placement1-decision-1
......
Status:
  Decisions:
    Cluster Name:  cluster1
    Reason:
Events:            <none>

Clean up

Undeploy placement controller from the cluster.

make undeploy-hub

Run e2e test cases as sanity check on an existing environment

In order to verify the Placement API on an existing environment with placement controller installed and well configured, you are able to run the e2e test cases as sanity check by following the steps below.

Build the binary of the e2e test cases

make build-e2e

And then run the e2e test cases against an existing environment.

./e2e.test --ginkgo.v --ginkgo.label-filter=sanity-check -hub-kubeconfig=/path/to/file

In an environment that has already had the global clusterset created, you can skip the creation of the global clusterset during testing.

./e2e.test --ginkgo.v --ginkgo.label-filter=sanity-check -hub-kubeconfig=/path/to/file -create-global-clusterset=false

Since the e2e test cases create fake ManagedClusters (without agent installed) during testing, in a full featured OCM environment (with registration controller running on the hub), a taint cluster.open-cluster-management.io/unreachable will be added to those fake ManagedClusters automatically. You have to tolerate this taint when running e2e test cases in such an environment.

./e2e.test --ginkgo.v --ginkgo.label-filter=sanity-check -hub-kubeconfig=/path/to/file -tolerate-unreachable-taint