In this brief demo we'd like to show easily you can bring up a cluster with cluster addons. What you will get is:
- a cluster with two nodes (one controlplane, one worker) using VMs
kubeadmbringing up the addon using the proposed addon installer- the gpu plugin deployed to the cluster as a daemonset
The Cluster Addons sub-project of SIG Cluster Lifecycle brought together people from various areas of the Kubernetes landscape and produced the following specifications and code:
- Addon Operators
- Addon Installer
- the installer itself
- KEP Installing Addons in Kubeadm
- Integration of addon installer in kubeadm
- Discussions about how to integrate into
kopsare underway.
Check out the meeting minutes (and videos) for more details and please join the next meetings.
First make sure you have go 1.13, git and bazel installed. We'll assume you run Ubuntu or Debian here for simplicity:
sudo apt install -y apt-transport-https git snapd curl docker.io vagrant vagrant-cachier virtualbox
sudo snap install go --channel 1.13/stable --classic
sudo snap install hub --classic
echo "deb [arch=amd64] https://storage.googleapis.com/bazel-apt stable jdk1.8" | sudo tee /etc/apt/sources.list.d/bazel.list
curl https://bazel.build/bazel-release.pub.gpg | sudo apt-key add -
sudo apt update
sudo apt install -y bazelWe are going to build a branch of kubeadm, which adds support for
addons.
How this can be used is described in the KEP.
export GOPATH=~/go/
mkdir -p $GOPATH/src/github.com/kubernetes
cd $GOPATH/src/github.com/kubernetes/
git clone --single-branch https://github.com/kubernetes/kubernetes.git --depth 1
cd kubernetes
hub pr checkout 85224
GO111MODULE=on bazel build //cmd/kubeadmThen get the source of this repo:
cd ~
git clone https://github.com/dholbach/gpu-demoAll you need to do for bring up is
cd gpu-demo
vagrant upFirst of all, vagrant will bring up two VMs using VirtualBox, it will
install a whole raft of dependencies and make scripts, config and our build
of kubeadm and the GPU plugin available in the VMs, the network is
configured as well.
Relevant code:
- https://github.com/dholbach/gpu-demo/blob/master/Vagrantfile
- https://github.com/dholbach/gpu-demo/blob/master/config/vm-prep.sh
In a second step, our build of kubeadm is used to bring up the
cluster.
Note: the relevant commands to bring up the cluster can be seen here.
All that is required to bring up the cluster is a simple cluster config that is passed to
kubeadm.Here are the addon related config bits.
To verify all is working as intended, all you need to do is:
$ vagrant ssh controlplane-1 -- -tt sudo bash
root@controlplane-1:~# kubectl get pods -n kube-system
NAME READY STATUS RESTARTS AGE
coredns-5644d7b6d9-fx27j 1/1 Running 0 5m35s
coredns-5644d7b6d9-r7w8f 1/1 Running 0 5m35s
etcd-controlplane-1 1/1 Running 0 4m46s
intel-gpu-plugin-ghsdk 1/1 Running 0 3m54s
intel-gpu-plugin-lmlv8 1/1 Running 0 5m4s
kube-apiserver-controlplane-1 1/1 Running 0 4m44s
kube-controller-manager-controlplane-1 1/1 Running 0 4m43s
kube-proxy-bpccd 1/1 Running 0 5m14s
kube-proxy-ncm8m 1/1 Running 0 5m36s
kube-scheduler-controlplane-1 1/1 Running 0 4m33s
weave-net-8td8g 2/2 Running 0 5m36s
weave-net-fwzbj 2/2 Running 1 5m14s
root@controlplane-1:~#