Technically Speaking

The Official Bigstep Blog


Kubernetes on Bare Metal Cloud

Kubernetes is gaining increasingly more ground in many companies and many applications. Most of you already use GCP and AWS to deploy a Kubernetes cluster as it’s stress-free and ready to use. But what if you wanted to customize your setup? What if you wanted to max out the security and keep it on premise? You can build your own Kubernetes cluster on bare metal. Or better, on a bare metal cloud.

In this tutorial, I present two options for you to deploy a Kubernetes cluster: one standalone node (for testing/developing environments) and a multi-node cluster (production-ready).

1. How to set up a standalone node on CentOS 7

For the first setup, I’ve used a bare metal instance of 8 cores and 32 GB RAM with SSD storage running CentOS 7.

The first step is to configure the repo:

cat <<EOF > /etc/yum.repos.d/kubernetes.repo

Disable SELinux:

setenforce 0
sed -i --follow-symlinks 's/SELINUX=enforcing/SELINUX=disabled/g' /etc/sysconfig/selinux

Setting SELinux in permissive mode by running setenforce 0 and sed effectively disables it. This is required to allow containers to access the host filesystem, which is needed, for example, by pod networks. You have to do this until SELinux support is improved in the kubelet.

Install packages:

yum install -y docker kubelet kubeadm kubectl 
systemctl enable docker
systemctl start docker
systemctl enable kubelet

Some users on RHEL/CentOS 7 have reported issues with traffic being routed incorrectly due to iptables being bypassed. You should ensure net.bridge.bridge-nf-call-iptables is set to 1 in your sysctl config, e.g.

cat <<EOF >  /etc/sysctl.d/k8s.conf
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-iptables = 1
sysctl -system

Make sure that the br_netfilter module is loaded before this step. This can be done by running:

lsmod | grep br_netfilter

In order to load it, explicitly call:

modprobe br_netfilter

Set swap to OFF:

swapoff -a

Check /etc/fstab file and comment out the swap mounting point

Initialize Kubernetes node:

kubeadm init --pod-network-cidr=

In this case, you don’t need to run the join command as you want a standalone server.

Since $HOME is /root/ in my case it looks something like:
mkdir /root/.kube
cp -i /etc/kubernetes/admin.conf /root/.kube/config

Implement networking model using Flannel:

kubectl apply -f

And make this node a standalone one:

kubectl taint nodes --all
kubectl get nodes

Ta-da! You now have a standalone node on CentOS 7!

2. How to set up a multi-node cluster on Ubuntu 18.04

For this production-ready environment, I’ve used 3 instances of 8 cores, 32GB RAM with SSD storage. For the master node, I have added another disk drive to use it as an NFS persistent storage.

For each node run the following:

Configure repo:

apt update && apt install -y apt-transport-https curl
curl -s | apt-key add -

cat <<EOF >/etc/apt/sources.list.d/kubernetes.list
deb kubernetes-xenial main
apt update

Disable swap:

swapoff -a
Check /etc/fstab file and comment out the swap mounting point

Install packages:

apt install -y kubelet kubeadm kubectl
systemctl enable docker

Make sure that the br_netfilter module is loaded before this step. This can be done by running:

lsmod | grep br_netfilter

In order to load it, explicitly call:

modprobe br_netfilter

Then initialize Kubernetes node on the master node:

kubeadm init --pod-network-cidr=

mkdir $HOME/.Kube
cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
kubectl apply -f

Join the other two working nodes to the master by running the command in the output of init:

kubeadm join --token wbktuv.tmnyrg4plxh0d4q8 \
    --discovery-token-ca-cert-hash sha256:d5ce6cb1517e96d4ecaf9cff7537c57a5bead7ce307833664304347092a3cf52

On the master node, check the nodes:

kubectl get nodes

Install a kubeadmin dashboard. If you are used to a UI, you can install a dashboard similar to the one Google uses.

kubectl apply -f

Edit the service in order to expose it. Change ClusterIP to NodePort:

kubectl --namespace kube-system edit service kubernetes-dashboard

Create a service account and get authentication token and create two yaml files accordingly:

vim dashboard-admin.yaml

kind: ClusterRoleBinding
  name: admin-user
  kind: ClusterRole
  name: cluster-admin
- kind: ServiceAccount
  name: admin-user
  namespace: kube-system

vim dasboard-adminuser.yaml

apiVersion: v1
kind: ServiceAccount
  name: admin-user
  namespace: kube-system

kubectl apply -f dashboard-admin.yaml

kubectl apply -f dashboard-adminuser.yaml

Get token:

Start kube proxy in a screen:

kubectl proxy

Connect to the UI dashboard using the node’s IP and external port:

[root@instance-57037 ~]# kubectl --namespace kube-system get service kubernetes-dashboard
NAME                   TYPE       CLUSTER-IP      EXTERNAL-IP   PORT(S)         AGE
kubernetes-dashboard   NodePort   <none>        443:30777/TCP   3m

https://your_nodes_IP:30777 and use the token from above.

This is it! You now have two options to build your own Kubernetes cluster and a way to use a dashboard to manage it by UI.

The main advantages of using Bigstep Bare Metal Cloud for deploying your customized Kubernetes cluster are:

  1. Having the performance of a bare metal and the flexibility of a cloud (you can upgrade the hardware specs anytime)
  2. Skipping the virtualization layer
  3. In case of hardware failure, you only lose the compute power, and not the data

However, keep in mind that:

  1. You have to manually install the cluster and customize it
  2. Depending on your security settings/protocols, you may encounter some setup issues regarding iptables

Please contact us by email if you have any questions or would like to suggest additional performance tweaks. Furthermore, if you are in need of bare metal cloud, you know where to find us.

About the author

Rares Cosereanu is one of Bigstep’s tech-savvy Senior Systems Administrators, huge fan of cloud architectures and open source containerized distributed solutions.

Got a question? Need advice? We're just one click away.
Sharing is caring:TwitterFacebookLinkedinPinterestEmail

Readers also enjoyed:

Nginx in WHM, what is the best option?

For a long time, the only web server available in WHM was Apache, with Litespeed as a paid premium option. However, Nginx is widely considered today to…

Using lsapi in WHM

An update for EasyApache 4 released in early February has introduced a version of mod_lsapi that can be installed by all WHM users. This high-performance…

Leave a Reply

Your email address will not be published.

* Required fields to post your comments.
Please review our Privacy Notice in order to understand how we process your personal data and what are your rights in this respect.