Rancher Kubernetes Engine with TungstenFabric on Ubuntu

Rancher Kubernetes Engine with TungstenFabric on Ubuntu.txt

1) Install Ubuntu on nodes and set hostname and IP on all nodes

2) Prepare nodes. Run these commands as the root user on all nodes
	
	#turn off swap
	swapoff -a
	
	#install packages
	apt-get install -y ntp docker.io

	#configure ntp. set pool.ntp.org to your local or preferred server
	sed -i '/pool/d' /etc/ntp.conf
	echo 'server pool.ntp.org iburst' | tee -a /etc/ntp.conf
	service ntp restart

	#disable firewall
	sudo ufw disable 

3) Distribute ssh keys

	#copy existing ssh key pair to /root/.ssh/ on the master node or create a new ssh key pair with 
	ssh-keygen -t rsa -b 4096 -C "your_email@domain.com"
	#distribute the public key to each node with
	ssh-copy-id user@node

4) Instantiate RKE Kubernetes cluster with no CNI plugin

	#run these commands as root on the master node
	
	#cd and download rke binary
	cd /root/ ; wget https://github.com/rancher/rke/releases/download/v0.3.0/rke_linux-amd64

	#rename, chmod, move, test rke binary
	mv rke_linux-amd64 rke ; chmod +x rke ; mv rke /usr/bin ; rke --version
	
	#configure the cluster.yml using 'rke config'. 
	#set the following variables for each node and accept the default for all other variables
	
	rke config
	
	#Number of Hosts  				#enter the number of nodes in the cluster
	#SSH Address of Hosts 				#enter IP address for node
	#SSH User of host 				#enter root
	#Is host a Control Plane host 			#enter y for master node
	#Is host a Worker Host 				#enter y for worker node 
	#Is host an etcd host  				#enter y for master node
	#Override Hostname of host 			#enter hostname of node 
	#Network Plugin type 				#enter none
	#Service Cluster IP Range                       #enter 10.96.0.0/12
	#Cluster Network CIDR                           #enter 10.32.0.0/12
	#Cluster DNS Service IP                         #enter 10.96.0.3

	#run rke up to instantiate rke kubernetes cluster
	rke up

5) Install TungstenFabric 

	#run these commands as root on the master node

	#install kubectl
	sudo snap install kubectl --classic 

	#copy kubeconfig into place
	mkdir /root/.kube ; cp /root/kube_config_cluster.yml /root/.kube/config
	
	#label master nodes with the 'node-role.kubernetes.io/master=true' label. 
	#replace <MASTER_NODE_NAME> with the hostname of the master node
	kubectl label nodes <MASTER_NODE_NAME> node-role.kubernetes.io/master=true

	#add /var/lib/contrail/ports/vm to rke kubelet container ON ALL NODES
	docker exec -it kubelet mkdir -p /var/lib/contrail/ports/vm
	
	#install TungstenFabric via manifest. set <MASTER_NODE_IP> to the master node ip
	export K8S_MASTER_IP=<MASTER_NODE_IP> ; CONTRAIL_REPO="docker.io\/opencontrailnightly"; CONTRAIL_RELEASE="latest"; mkdir -pm 777 /var/lib/contrail/kafka-logs; curl https://raw.githubusercontent.com/Juniper/contrail-kubernetes-docs/master/install/kubernetes/templates/contrail-single-step-cni-install-ubuntu.yaml | sed "s/{{ K8S_MASTER_IP }}/$K8S_MASTER_IP/g; s/{{ CONTRAIL_REPO }}/$CONTRAIL_REPO/g; s/{{ CONTRAIL_RELEASE }}/$CONTRAIL_RELEASE/g" >> /root/tf.yml ; kubectl apply -f /root/tf.yml

	#fix coredns
	kubectl get configmap -nkube-system coredns -o yaml >> /root/coredns-configmap.yaml
	export coredns_line="        forward . 10.47.255.253"
  	sed -i "s/.*forward.*/$coredns_line/" /root/coredns-configmap.yaml
        kubectl get deployment -nkube-system coredns -o yaml >> /root/coredns-deployment.yaml
	sed -i "/        livenessProbe:/,+9d" /root/coredns-deployment.yaml 
	sed -i "/        readinessProbe:/,+8d" /root/coredns-deployment.yaml
	kubectl apply -f /root/coredns-configmap.yaml
	kubectl apply -f /root/coredns-deployment.yaml

	#get pod status 
	kubectl get pods --all-namespaces -owide
	
	#for now it is safe to ignore analytics db crashloop
	#kube-system     contrail-analyticsdb-r6474                2/3     CrashLoopBackOff 	
		
	#get TungstenFabric status once all TungstenFabric pods are running
	contrail-status

	login to webui at https://<MASTER_NODE_IP>:8143 u: admin p: contrail123