Cluster API bootstrap provider k3s (CABP3) is a component of Cluster API that is responsible for generating a cloud-init script to turn a Machine into a Kubernetes Node; this implementation brings up k3s clusters instead of full kubernetes clusters.
CABP3 is the bootstrap component of Cluster API for k3s and brings in the following CRDS and controllers:
- k3s bootstrap provider (KThrees, KThreesTemplate)
Cluster API ControlPlane provider k3s (CACP3) is a component of Cluster API that is responsible for managing the lifecycle of control plane machines for k3s; this implementation brings up k3s clusters instead of full kubernetes clusters.
CACP3 is the controlplane component of Cluster API for k3s and brings in the following CRDS and controllers:
- k3s controlplane provider (KThreesControlPlane)
Warning: Project and documentation are in an early stage, there is an assumption that an user of this provider is already familiar with ClusterAPI.
Check out the ClusterAPI Quickstart page to see the prerequisites for ClusterAPI.
Three main pieces are
- Bootstrap cluster. In the
samples/azure/azure-setup.shscript, I use k3d, but feel free to use kind as well. - clusterctl. Please check out ClusterAPI Quickstart for instructions.
- Azure Service Principals. For more information go to CAPZ Getting Started
CABP3 has been tested only on with an Azure and AzureStackHCI environment. To try out the Azure flow, fork the repo and look at samples/azure/azure-setup.sh.
CACP3 is alive! Sample now includes the K3s Control Plane Provider. If you run the sample script you will get a cluster with a control plane and two workers.
Then run the following to scale the control plane...
kubectl scale kthreescontrolplane ${CLUSTER_NAME}-control-plane --replicas 3- Support for External Databases
- Fix Token Logic
- Setup CAPA and CAPV samples
- Post an issue!
Steps to run it in a VSphere environgment(validated with tkg 1.3.1)
- Create a tkg management cluster
- change the BOOTSTRAP_IMG and CONTROLPLANE_IMG in the Makefile to needed image name and then do make docker-build-bootstrap make docker-build-controlplane and push the docker images to your repo.
- goto smaples/vsphere/ directory
- export CLUSTER_NAME=<> with your needed workload k3s cluster name
- cat k3s-template-master-only.yaml | envsubst > k3s-master-only.yaml
- cat k3s-template-master-only-addons.yaml | envsubst > k3s-master-only.yaml
- copy the k3s-master-only.yaml and k3s-master-only.yaml to your tkg management cluster
- make changes in the yaml file with vsphere /nsx credentials etc and also vip needed for your cluster.
- kubectl create namespace ${CLUSTER_NAME} #same value as cluster name
- kubectl create -f k3s-master-only.yaml && kubectl create -f k3s-master-only.yaml
- after the ndoe is created, login using the credentials capv/VMware1! and run following command sudo k3s kubectl label node --all node-role.kubernetes.io/master="" --overwrite=true This step is needed for ccm to be installed using kapp so that node can be untainted and other compoennts/pods can be installed. so without this step, no pods will be run.