Author: Paulo Pires [email protected]
This document describes, step-by-step, how to provision a Kubernetes cluster comprised of:
- One Linux machine acting as Kubernetes master node and OVN central database.
- One Linux machine acting as Kubernetes worker node.
- One Windows machine acting as Kubernetes worker node.
Many thanks to the great people that helped achieve this, namely:
- Alin Serdean (Cloudbase Solutions)
- Alin Balutoiu (Cloudbase Solutions)
- Feng Min (Google)
- Peter Hornyack (Google)
- Use Google Cloud Platform (GCP) , namely Google Compute Engine (GCE) VMs.
gcloud
CLI tool
- Linux machines(s) run Ubuntu 16.04 with latest updates.
- Windows machine(s) run Windows Server 2016 with latest updates.
- Administrator access to all VMs, i.e.
root
in Linux machines.
Let's provision the master VM:
gcloud compute instances create "sig-windows-master" \
--zone "us-east1-d" \
--machine-type "custom-2-2048" \
--subnet "default" \
--can-ip-forward \
--maintenance-policy "MIGRATE" \
--tags "http-server","https-server" \
--image "ubuntu-1604-xenial-v20170125" \
--image-project "ubuntu-os-cloud" \
--boot-disk-size "50" \
--boot-disk-type "pd-ssd" \
--boot-disk-device-name "sig-windows-master"
When it's ready, SSH into it:
gcloud compute ssh --zone "us-east1-d" "sig-windows-master"
ATTENTION: From now on, it's assumed you're logged-in as root
.
Since we'll need Docker, let's install it:
curl -fsSL https://yum.dockerproject.org/gpg | apt-key add -
echo "deb https://apt.dockerproject.org/repo ubuntu-xenial main" > sudo tee /etc/apt/sources.list.d/docker.list
apt-get update
apt-get install -y docker.io
We also need to download the contents of this repository which will be used shortly:
cd ~
git clone https://github.com/apprenda/kubernetes-ovn-heterogeneous-cluster
cd kubernetes-ovn-heterogeneous-cluster/deb
dpkg -i openvswitch-common_2.6.2-1_amd64.deb \
openvswitch-datapath-dkms_2.6.2-1_all.deb \
openvswitch-switch_2.6.2-1_amd64.deb \
ovn-common_2.6.2-1_amd64.deb \
ovn-central_2.6.2-1_amd64.deb \
ovn-docker_2.6.2-1_amd64.deb \
ovn-host_2.6.2-1_amd64.deb \
python-openvswitch_2.6.2-1_all.deb
Finally, reboot:
reboot
SSH again into the machine and let's proceed.
TODO is this still needed after the reboot?
rmmod openvswitch
insmod /lib/modules/$(uname -r)/updates/dkms/openvswitch.ko
modprobe vport-geneve
Create the OVS bridge interface:
ovs-vsctl add-br br-int
export TUNNEL_MODE=geneve
export LOCAL_IP=10.142.0.2
export MASTER_IP=10.142.0.2
ovs-vsctl set Open_vSwitch . external_ids:ovn-remote="tcp:$MASTER_IP:6642" \
external_ids:ovn-nb="tcp:$MASTER_IP:6641" \
external_ids:ovn-encap-ip="$LOCAL_IP" \
external_ids:ovn-encap-type="$TUNNEL_MODE"
ovs-vsctl get Open_vSwitch . external_ids
You should see something like:
{hostname=sig-windows-master.c.apprenda-project-one.internal, "k8s-api-server"="127.0.0.1:8080", ovn-encap-ip="10.142.0.2", ovn-encap-type=geneve, ovn-nb="tcp:10.142.0.2:6641", ovn-remote="tcp:10.142.0.2:6642", system-id="e7af27f6-a218-40bb-8d4f-af67600abd17"}
We are now ready to set-up Kubernetes master node.
ATTENTION: From now on, it's assumed you're logged-in as root
.
First, we need to run etcd
, the store used by Kubernetes. The following is a one-time step only:
cd ~/kubernetes-ovn-heterogeneous-cluster/master
cp -R systemd/*.service /etc/systemd/system/
systemctl daemon-reload
systemctl enable etcd3
systemctl start etcd3
Then, proceed to set-up Kubernetes:
./make-certs
mkdir -p /etc/kubernetes
cp -R manifests /etc/kubernetes/
systemctl enable kubelet
systemctl start kubelet
curl -Lskj -o /usr/bin/kubectl https://storage.googleapis.com/kubernetes-release/release/v1.5.2/bin/linux/amd64/kubectl
chmod +x /usr/bin/kubectl
kubectl config set-cluster default-cluster --server=https://10.142.0.2 --certificate-authority=/etc/kubernetes/tls/ca.pem
kubectl config set-credentials default-admin --certificate-authority=/etc/kubernetes/tls/ca.pem --client-key=/etc/kubernetes/tls/admin-key.pem --client-certificate=/etc/kubernetes/tls/admin.pem
kubectl config set-context local --cluster=default-cluster --user=default-admin
kubectl config use-context local
kubectl create -f ../kubedns-deployment.yaml
kubectl create -f ../kubedns-service.yaml
Last step is to configure pod networking for this node:
ovs-vsctl set Open_vSwitch . external_ids:k8s-api-server="127.0.0.1:8080"
apt-get install -y python-pip
cd ~
git clone https://github.com/openvswitch/ovn-kubernetes
cd ovn-kubernetes
pip install --prefix=/usr/local .
ovn-k8s-overlay master-init \
--cluster-ip-subnet="10.244.0.0/16" \
--master-switch-subnet="10.244.1.0/24" \
--node-name="10.142.0.2"
systemctl enable ovn-k8s-watcher
systemctl start ovn-k8s-watcher
By this time, your master node is ready:
kubectl get nodes
kubectl -n kube-system get pods
Let's proceed to set-up the worker nodes.
Let's provision the master VM:
gcloud compute instances create "sig-windows-worker-linux" \
--zone "us-east1-d" \
--machine-type "custom-2-2048" \
--subnet "default" \
--can-ip-forward \
--maintenance-policy "MIGRATE" \
--image "ubuntu-1604-xenial-v20170125" \
--image-project "ubuntu-os-cloud" \
--boot-disk-size "50" \
--boot-disk-type "pd-ssd" \
--boot-disk-device-name "sig-windows-worker-linux"
When it's ready, SSH into it:
gcloud compute ssh --zone "us-east1-d" "sig-windows-worker-linux"
ATTENTION: From now on, it's assumed you're logged-in as root
.
Since we'll need Docker, let's install it:
curl -fsSL https://yum.dockerproject.org/gpg | apt-key add -
echo "deb https://apt.dockerproject.org/repo ubuntu-xenial main" > sudo tee /etc/apt/sources.list.d/docker.list
apt-get update
apt-get install -y docker.io
We also need to download the contents of this repository which will be used shortly:
cd ~
git clone https://github.com/apprenda/kubernetes-ovn-heterogeneous-cluster
cd kubernetes-ovn-heterogeneous-cluster/deb
dpkg -i openvswitch-common_2.6.2-1_amd64.deb \
openvswitch-datapath-dkms_2.6.2-1_all.deb \
openvswitch-switch_2.6.2-1_amd64.deb \
ovn-common_2.6.2-1_amd64.deb \
ovn-central_2.6.2-1_amd64.deb \
ovn-docker_2.6.2-1_amd64.deb \
ovn-host_2.6.2-1_amd64.deb \
python-openvswitch_2.6.2-1_all.deb
Finally, reboot:
reboot
SSH again into the machine and let's proceed.
TODO is this still needed after the reboot?
rmmod openvswitch
insmod /lib/modules/$(uname -r)/updates/dkms/openvswitch.ko
modprobe vport-geneve
Create the OVS bridge interface:
ovs-vsctl add-br br-int
export TUNNEL_MODE=geneve
export LOCAL_IP=10.142.0.3
export MASTER_IP=10.142.0.2
ovs-vsctl set Open_vSwitch . external_ids:ovn-remote="tcp:$MASTER_IP:6642" \
external_ids:ovn-nb="tcp:$MASTER_IP:6641" \
external_ids:ovn-encap-ip="$LOCAL_IP" \
external_ids:ovn-encap-type="$TUNNEL_MODE"
ovs-vsctl get Open_vSwitch . external_ids
Then, proceed to set-up Kubernetes:
cd ~/kubernetes-ovn-heterogeneous-cluster/node
# TODO copy the two CA files below from master
chmod 600 /etc/kubernetes/tls/ca-key.pem
chmod 660 /etc/kubernetes/tls/ca.pem
chgrp kube-cert /etc/kubernetes/*
./linux-make-certs
mkdir -p /etc/kubernetes
cp -R manifests /etc/kubernetes/
cp kubeconfig.yaml /etc/kubernetes/
cp -R systemd/*.service /etc/systemd/system/
systemctl daemon-reload
systemctl enable kubelet
systemctl start kubelet
ovs-vsctl set Open_vSwitch . external_ids:k8s-api-server="10.142.0.2:8080"
ovs-vsctl set Open_vSwitch . \
external_ids:k8s-api-server="https://10.142.0.2" \
external_ids:k8s-ca-certificate="/etc/kubernetes/tls/ca.pem" \
external_ids:k8s-api-token="/etc/kubernetes/tls/node.pem"
mkdir -p /opt/cni/bin && cd /opt/cni/bin
curl -Lskj -o cni.tar.gz https://github.com/containernetworking/cni/releases/download/v0.4.0/cni-v0.4.0.tgz
tar zxf cni.tar.gz
rm -f cni.tar.gz
apt-get install -y python-pip
git clone https://github.com/openvswitch/ovn-kubernetes
cd ovn-kubernetes
pip install --prefix=/usr/local .
ovn-k8s-overlay minion-init \
--cluster-ip-subnet="10.244.0.0/16" \
--minion-switch-subnet="10.244.2.0/24" \
--node-name="10.142.0.3"
By this time, your Linux worker node is ready:
kubectl get nodes
kubectl -n kube-system get pods
Let's proceed to setup the Windows worker node.
TODO
As root
, run:
apt-get update
apt-get install -y build-essential fakeroot dkms \
autoconf automake debhelper dh-autoreconf libssl-dev libtool \
python-all python-twisted-conch python-zopeinterface \
graphviz
cd ~
git clone https://github.com/openvswitch/ovs.git
cd ovs
git checkout branch-2.6
dpkg-checkbuilddeps
DEB_BUILD_OPTIONS='nocheck' fakeroot debian/rules binary
- Copy CA files from master.
-
etcd
container should use a host path for storing data, in order to survive restarts. - Linux worker node doesn't need the same OVS/OVN packages installed as the master node.
- Automate certificate generation based on node type and IP.
- Template all the things.
- Add other cloud providers documentation, e.g. AWS.
- Add gateway node instructions for enabling pod containers Internet access.
- Setup OVS TLS.