Adding KubeStellar to OCM#
This page shows one concrete example of adding KubeStellar to an existing OCM system. In particular, a hub plus two managed clusters almost exactly as created by the OCM Quick Start. In terms of the full Installation and Usage outline of KubeStellar, the modified OCM Quick Start has already: established some, but not all, of the software prerequisites; acquired the ability to use a Kube cluster as KubeFlex hosting cluster; created an Inventory and Transport Space; created two Workload Execution Clusters (WECs) and registered them. These are the boxes outlined in red in the following flowchart.
.
- Setup
- Install remaining software prerequisites
- Cleanup from previous runs
- OCM Quick Start with Ingress
- Label WECs for selection by examples
- Install Kubestellar core components
- Exercise KubeStellar
- Troubleshooting
Setup#
Continuing with the spirit of the OCM Quick Start, this is one way to produce a very simple system --- suitable for study but not production usage. For general setup information, see the full story.
Install software prerequisites#
The following command will check for the prerequisites that you will need for the later steps. See the prerequisites doc for more details.
bash <(curl https://raw.githubusercontent.com/kubestellar/kubestellar/v0.25.0-rc.2/hack/check_pre_req.sh) kflex ocm helm kubectl docker kind
Set the Version appropriately as an environment variable#
OCM Quick Start with Ingress#
This recipe uses a modified version of the OCM Quick Start. The modification is because KubeStellar needs the hosting cluster to have an Ingress controller with SSL passthrough enabled. The modified Quick Start script has the following modifications compared to the baseline.
- The
kind
cluster created for the hub has an additional port mapping, where the Ingress controller listens. - The script installs the NGINX Ingress Controller into the hub cluster, then patches the controller to enable SSL passthrough, and later waits for it to be in service.
You can invoke the modified OCM Quick Start as follows.
curl -L https://raw.githubusercontent.com/kubestellar/kubestellar/refs/heads/0.25.0-rc.2/scripts/ocm-local-up-for-ingress.sh | bash
Like the baseline, this script creates a kind
cluster named "hub" to
serve as hub cluster (known in KubeStellar as an Inventory and
Transport Space, ITS) and two kind
clusters named "cluster1" and
"cluster2" to serve as managed clusters (known in KubeStellar as
Workload Execution Clusters, WECs), and registers them in the hub.
Label WECs for selection by examples#
The examples will use label selectors to direct workload to WECs (in terms of their ManagedCluster representations). The following commands apply labels that will be used in the examples.
kubectl --context kind-hub label managedcluster cluster1 location-group=edge name=cluster1
kubectl --context kind-hub label managedcluster cluster2 location-group=edge name=cluster2
Use Core Helm chart to initialize KubeFlex, recognize ITS, and create WDS#
This chart instance will do the following.
- Install KubeFlex in the hosting cluster.
- Assign the hosting cluster the role of an ITS, named "its1".
- Create a KubeFlex ControlPlane named "wds1" to play the role of a WDS.
- Install the KubeStellar core stuff in the hosting cluster.
helm --kube-context kind-hub upgrade --install ks-core oci://ghcr.io/kubestellar/kubestellar/core-chart \
--version $kubestellar_version \
--set-json='ITSes=[{"name":"its1", "type":"host"}]' \
--set-json='WDSes=[{"name":"wds1"}]' \
--set-json='verbosity.default=5' # so we can debug your problem reports
That command will print some notes about how to get a kubeconfig "context" named "wds1" defined. Do that, because this context is used in the steps that follow. The notes assume that your current kubeconfig context is the one where the Helm chart was installed, which is not necessarily true --- so take care for that too.
kubectl config use-context kind-hub
kflex ctx --set-current-for-hosting # make sure the KubeFlex CLI's hidden state is right for what the Helm chart just did
kflex ctx --overwrite-existing-context wds1
For more information about this Helm chart, see its documentation.
Exercise KubeStellar#
Use the following commands to wait for the KubeStellar core Helm chart to finish setting up the WDS, because the examples assume that this has completed.
while [ -z "$(kubectl --context wds1 get crd bindingpolicies.control.kubestellar.io --no-headers -o name 2> /dev/null)" ] ; do
sleep 5
done
kubectl --context wds1 wait --for condition=Established crd bindingpolicies.control.kubestellar.io
Proceed to Scenario 1 (multi-cluster workload deployment with kubectl) in the example scenarios after defining the shell variables that characterize the setup done above. Following are the settings for those variables, whose meanings are defined at the start of the example scenarios document.
host_context=kind-hub
its_cp=its1
its_context=${host_context}
wds_cp=wds1
wds_context=wds1
wec1_name=cluster1
wec2_name=cluster2
wec1_context=kind-$wec1_name
wec2_context=kind-$wec2_name
label_query_both=location-group=edge
label_query_one=name=cluster1
Troubleshooting#
In the event something goes wrong, check out the troubleshooting page to see if someone else has experienced the same thing