Easily Deploy CoreOS + Kubernetes Cluster to GCE

With a few simple scripts on your Mac OS X or Linux computer, you can deploy an elastic Kubernetes cluster on top of CoreOS using fleet to GCE. By default it is set to one master + three nodes.

Install dependencies if you do not have them on your OS X/Linux:

• You need Google Cloud account and GC SDK installed
• git

Clone this project and set settings:

git clone https://github.com/rimusz/coreos-multi-node-k8s-gce
cd coreos-multi-node-k8s-gce

• edit settings and set project and zone, the rest of settings you can adjust by your requirements if you need to.

Start ssh-agent as it is used extensively by this script

Confirm your ssh-agent is running. Run the following command. If you see the error message then it is not running.

$ ssh-add -l
Could not open a connection to your authentication agent.

Start it up. Run the command ssh-agent. You should see somthing simlar to the following.

$ ssh-agent
SSH_AUTH_SOCK=/tmp/ssh-ztSdihLxClVL/agent.24475; export SSH_AUTH_SOCK;
SSH_AGENT_PID=24476; export SSH_AGENT_PID;
echo Agent pid 24476;

Cut and paste the three lines the command output and paste back into the terminal you're using to run the kubernetes commnands. We can then do a quick test to confirm that we're running ssh-agent as we expect.

$ ssh-add -l
The agent has no identities.

ssh-agent is a whole other category of software. You can look more into it here.

Bootstrap Kubernetes Cluster and install local clients

• To bootstrap CoreOS cluster in GCE run:

1-bootstrap_cluster.sh

• To install local etcdctl, fleetctl and kubectl clients run:

2-get_k8s_fleet_etcd.sh

• Setup Kubernetes on CoreOS cluster run:

3-install_k8s_fleet_units.sh

And that's it, you now have Kubernetes cluster with one master + 3 nodes running in GCE and required OS X/Linux clients etcdctl, fleetctl and kubectl installed on your computer.

Usage

When you are done the bootstraping Kubernetes cluster, from the same folder run set_k8s_access.sh to get shell preset to work with etcd, fleet and Kubernetes master.

Script output will show the following:

/registry
/coreos.com

UNIT								MACHINE					ACTIVE		SUB
kube-apiserver.service			cc124065.../10.240.64.180	active	running
kube-controller-manager.service	cc124065.../10.240.64.180	active	running
kube-kubelet.service			21ed373b.../10.240.189.83	active	running
kube-kubelet.service			770ff9fd.../10.240.8.219	active	running
kube-kubelet.service			a9b4be28.../10.240.252.226	active	running
kube-proxy.service				21ed373b.../10.240.189.83	active	running
kube-proxy.service				770ff9fd.../10.240.8.219	active	running
kube-proxy.service				a9b4be28.../10.240.252.226	active	running
kube-scheduler.service			cc124065.../10.240.64.180	active	running

NAME             LABELS                                  STATUS
10.240.189.83    kubernetes.io/hostname=10.240.189.83    Ready
10.240.252.226   kubernetes.io/hostname=10.240.252.226   Ready
10.240.8.219     kubernetes.io/hostname=10.240.8.219     Ready

Type exit when you are finished ...

At this point you are ready to start playing with Kubernetes using the kubectl command tool.

• When you are done with set_k8s_access.sh just type exit or ctrl+d and on exit script will close all ssh connections to remote etcd control and Kubernetes master.

You can manually run 2-get_k8s_fleet_etcd.sh script to update OS X/Linux etcdctl, fleetctl and kubectl clients.

Adding and removing machines

To add more nodes, just update settings node_count and run 1-bootstrap_k8s_cluster.sh again. For removing nodes, simply destroy them via GCE developer console and fleet will reschedule the kube-kubelet and kube-proxy units.

If you are OS X user:

• Kubernetes Solo VM App can be found here Kubernetes Solo for OS X.
• Kubernetes Cluster VM App can be found here Kubernetes Cluster for OS X.